TADPOLE_BSWiMS

1 TADPOLE and BSWiMS

1.0.1 Loading the libraries

library("FRESA.CAD")
library(readxl)
library(igraph)
op <- par(no.readonly = TRUE)
pander::panderOptions('digits', 3)
pander::panderOptions('table.split.table', 400)
pander::panderOptions('keep.trailing.zeros',TRUE)

1.1 The data set

TADPOLE_D1_D2 <- read.csv("~/GitHub/BSWiMS/Data/TADPOLE/TADPOLE_D1_D2.csv")
TADPOLE_D1_D2_Dict <- read.csv("~/GitHub/BSWiMS/Data/TADPOLE/TADPOLE_D1_D2_Dict.csv")
TADPOLE_D1_D2_Dict_LR <- as.data.frame(read_excel("~/GitHub/BSWiMS/Data/TADPOLE/TADPOLE_D1_D2_Dict_LR.xlsx",sheet = "LeftRightFeatures"))


rownames(TADPOLE_D1_D2_Dict) <- TADPOLE_D1_D2_Dict$FLDNAME

1.2 Conditioning the data

# mm3 to mm
isVolume <- c("Ventricles","Hippocampus","WholeBrain","Entorhinal","Fusiform","MidTemp","ICV",
              TADPOLE_D1_D2_Dict$FLDNAME[str_detect(TADPOLE_D1_D2_Dict$TEXT,"Volume")]
              )


#TADPOLE_D1_D2[,isVolume] <- apply(TADPOLE_D1_D2[,isVolume],2,'^',(1/3))
TADPOLE_D1_D2[,isVolume] <- TADPOLE_D1_D2[,isVolume]^(1/3)

# mm2 to mm
isArea <- TADPOLE_D1_D2_Dict$FLDNAME[str_detect(TADPOLE_D1_D2_Dict$TEXT,"Area")]
TADPOLE_D1_D2[,isArea] <- sqrt(TADPOLE_D1_D2[,isArea])

# Get only cross sectional measurements
FreeSurfersetCross <- str_detect(colnames(TADPOLE_D1_D2),"UCSFFSX")

# The subset of baseline measurements
baselineTadpole <- subset(TADPOLE_D1_D2,VISCODE=="bl")
table(baselineTadpole$DX)

                   Dementia Dementia to MCI             MCI MCI to Dementia 
          7             336               1             864               5 
  MCI to NL              NL       NL to MCI 
          2             521               1 

rownames(baselineTadpole) <- baselineTadpole$PTID


validBaselineTadpole <- cbind(DX=baselineTadpole$DX,
                                 AGE=baselineTadpole$AGE,
                                 Gender=1*(baselineTadpole$PTGENDER=="Female"),
                                 ADAS11=baselineTadpole$ADAS11,
                                 ADAS13=baselineTadpole$ADAS13,
                                 MMSE=baselineTadpole$MMSE,
                                 RAVLT_immediate=baselineTadpole$RAVLT_immediate,
                                 RAVLT_learning=baselineTadpole$RAVLT_learning,
                                 RAVLT_forgetting=baselineTadpole$RAVLT_forgetting,
                                 RAVLT_perc_forgetting=baselineTadpole$RAVLT_perc_forgetting,
                                 FAQ=baselineTadpole$FAQ,
                                 Ventricles=baselineTadpole$Ventricles,
                                 Hippocampus=baselineTadpole$Hippocampus,
                                 WholeBrain=baselineTadpole$WholeBrain,
                                 Entorhinal=baselineTadpole$Entorhinal,
                                 Fusiform=baselineTadpole$Fusiform,
                                 MidTemp=baselineTadpole$MidTemp,
                                 ICV=baselineTadpole$ICV,
                                 baselineTadpole[,FreeSurfersetCross])


LeftFields <- TADPOLE_D1_D2_Dict_LR$LFN
names(LeftFields) <- LeftFields
LeftFields <- LeftFields[LeftFields %in% colnames(validBaselineTadpole)]
RightFields <- TADPOLE_D1_D2_Dict_LR$RFN
names(RightFields) <- RightFields
RightFields <- RightFields[RightFields %in% colnames(validBaselineTadpole)]

## Normalize to ICV
validBaselineTadpole$Ventricles=validBaselineTadpole$Ventricles/validBaselineTadpole$ICV
validBaselineTadpole$Hippocampus=validBaselineTadpole$Hippocampus/validBaselineTadpole$ICV
validBaselineTadpole$WholeBrain=validBaselineTadpole$WholeBrain/validBaselineTadpole$ICV
validBaselineTadpole$Entorhinal=validBaselineTadpole$Entorhinal/validBaselineTadpole$ICV
validBaselineTadpole$Fusiform=validBaselineTadpole$Fusiform/validBaselineTadpole$ICV
validBaselineTadpole$MidTemp=validBaselineTadpole$MidTemp/validBaselineTadpole$ICV

leftData <- validBaselineTadpole[,LeftFields]/validBaselineTadpole$ICV
RightData <- validBaselineTadpole[,RightFields]/validBaselineTadpole$ICV

## get mean and relative difference 
meanLeftRight <- (leftData + RightData)/2
difLeftRight <- abs(leftData - RightData)
reldifLeftRight <- difLeftRight/meanLeftRight
colnames(meanLeftRight) <- paste("M",colnames(meanLeftRight),sep="_")
colnames(difLeftRight) <- paste("D",colnames(difLeftRight),sep="_")
colnames(reldifLeftRight) <- paste("RD",colnames(reldifLeftRight),sep="_")


validBaselineTadpole <- validBaselineTadpole[,!(colnames(validBaselineTadpole) %in% 
                                               c(LeftFields,RightFields))]
validBaselineTadpole <- cbind(validBaselineTadpole,meanLeftRight,reldifLeftRight)

## Remove columns with too many NA more than %15 of NA
nacount <- apply(is.na(validBaselineTadpole),2,sum)/nrow(validBaselineTadpole) < 0.15
diagnose <- validBaselineTadpole$DX
pander::pander(table(diagnose))

	Dementia	Dementia to MCI	MCI	MCI to Dementia	MCI to NL	NL	NL to MCI
7	336	1	864	5	2	521	1

validBaselineTadpole <- validBaselineTadpole[,nacount]
## Remove character columns
ischar <- sapply(validBaselineTadpole,class) == "character"
validBaselineTadpole <- validBaselineTadpole[,!ischar]
## Place back diagnose
validBaselineTadpole$DX <- diagnose


validBaselineTadpole <- validBaselineTadpole[complete.cases(validBaselineTadpole),]
ischar <- sapply(validBaselineTadpole,class) == "character"
validBaselineTadpole[,!ischar] <- sapply(validBaselineTadpole[,!ischar],as.numeric)

colnames(validBaselineTadpole) <- str_remove_all(colnames(validBaselineTadpole),"_UCSFFSX_11_02_15_UCSFFSX51_08_01_16")
colnames(validBaselineTadpole) <- str_replace_all(colnames(validBaselineTadpole)," ","_")
validBaselineTadpole$LONISID <- NULL
validBaselineTadpole$IMAGEUID <- NULL
validBaselineTadpole$LONIUID <- NULL

diagnose <- as.character(validBaselineTadpole$DX)
validBaselineTadpole$DX <- diagnose
pander::pander(table(validBaselineTadpole$DX))

Dementia	Dementia to MCI	MCI	MCI to Dementia	MCI to NL	NL	NL to MCI
244	1	711	2	2	452	1

validDX <- c("NL","MCI","Dementia")

validBaselineTadpole <- validBaselineTadpole[validBaselineTadpole$DX %in% validDX,]
validBaselineTadpole$DX <- as.factor(validBaselineTadpole$DX)
pander::pander(table(validBaselineTadpole$DX))

Dementia	MCI	NL
244	711	452

1.3 Get the Time To Event on MCI Subjects

subjectsID <- rownames(validBaselineTadpole)
visitsID <- unique(TADPOLE_D1_D2$VISCODE)
baseDx <- TADPOLE_D1_D2[TADPOLE_D1_D2$VISCODE=="bl",c("PTID","DX","EXAMDATE")]
rownames(baseDx) <- baseDx$PTID 
baseDx <- baseDx[subjectsID,]
lastDx <- baseDx
toDementia <- baseDx
table(lastDx$DX)

Dementia MCI NL 244 711 452

hasDementia <- lastDx$PTID[str_detect(lastDx$DX,"Dementia")]


for (vid in visitsID)
{
  DxValue <- TADPOLE_D1_D2[TADPOLE_D1_D2$VISCODE==vid,c("PTID","DX","EXAMDATE")]
  rownames(DxValue) <- DxValue$PTID 
  DxValue <- DxValue[DxValue$PTID %in% subjectsID,]
  noDX <- DxValue$PTID[nchar(DxValue$DX) < 1]
  print(length(noDX))
  DxValue[noDX,] <- lastDx[noDX,]
  inLast <- lastDx$PTID[lastDx$PTID %in% DxValue$PTID]
  print(length(inLast))
  lastDx[inLast,] <- DxValue[inLast,]
  noDementia <- !(toDementia$PTID %in% hasDementia)
  toDementia[noDementia,] <- lastDx[noDementia,]
  hasDementia <- unique(c(hasDementia,lastDx$PTID[str_detect(lastDx$DX,"Dementia")]))
}

[1] 0 [1] 1407 [1] 2 [1] 1320 [1] 6 [1] 1212 [1] 23 [1] 1090 [1] 802 [1] 1054 [1] 29 [1] 706 [1] 20 [1] 212 [1] 14 [1] 167 [1] 32 [1] 551 [1] 25 [1] 297 [1] 18 [1] 130 [1] 665 [1] 665 [1] 112 [1] 112 [1] 176 [1] 176 [1] 177 [1] 177 [1] 624 [1] 624 [1] 251 [1] 251 [1] 159 [1] 159 [1] 7 [1] 7 [1] 17 [1] 99 [1] 9 [1] 63 [1] 1 [1] 1

table(lastDx$DX)

   Dementia Dementia to MCI             MCI MCI to Dementia       MCI to NL 
        426               2             460              80               7 
         NL  NL to Dementia       NL to MCI 
        405               1              26 

baseMCI <-baseDx$PTID[baseDx$DX == "MCI"]
lastDementia <- lastDx$PTID[str_detect(lastDx$DX,"Dementia")]
lastDementia2 <- toDementia$PTID[str_detect(toDementia$DX,"Dementia")]
lastNL <- lastDx$PTID[str_detect(lastDx$DX,"NL")]

MCIatBaseline <- baseDx[baseMCI,]
MCIatEvent <- toDementia[baseMCI,]
MCIatLast <- lastDx[baseMCI,]

MCIconverters <- MCIatBaseline[baseMCI %in% lastDementia,]
MCI_No_converters <- MCIatBaseline[!(baseMCI %in% MCIconverters$PTID),]
MCIconverters$TimeToEvent <- (as.Date(toDementia[MCIconverters$PTID,"EXAMDATE"]) 
                                   - as.Date(MCIconverters$EXAMDATE))

sum(MCIconverters$TimeToEvent ==0)

[1] 0

MCIconverters$AtEventDX <- MCIatEvent[MCIconverters$PTID,"DX"]
MCIconverters$LastDX <- MCIatLast[MCIconverters$PTID,"DX"]

MCI_No_converters$TimeToEvent <- (as.Date(lastDx[MCI_No_converters$PTID,"EXAMDATE"]) 
                                   - as.Date(MCI_No_converters$EXAMDATE))

MCI_No_converters$LastDX <- MCIatLast[MCI_No_converters$PTID,"DX"]

MCI_No_converters <- subset(MCI_No_converters,TimeToEvent > 0)

1.3.1 Correlation Matrix Data

The heat map of the testing set.

cormat <- cor(validBaselineTadpole[,colnames(validBaselineTadpole) != "DX"],method="spearman")
diag(cormat) <- 0;
corrmax <- apply(cormat,2,max)
whomax <- colnames(cormat)[corrmax>0.75]
gplots::heatmap.2(abs(cormat[whomax,whomax]),
                  trace = "none",
                  scale = "none",
                  mar = c(10,10),
                  col=rev(heat.colors(5)),
                  main = "Baseline TADPOLE Correlation",
                  cexRow = 0.5,
                  cexCol = 0.5,
                  key.title=NA,
                  key.xlab="Spearman Correlation",
                  xlab="Feature", ylab="Feature")

par(op)

2 Predicting ADAS13

Here we will diagnose ADAS13

2.1 Training and testing sets

TrainFraction <- 0.5;

TADPOLECrossMRI <- validBaselineTadpole
summary(TADPOLECrossMRI$ADAS13)

Min. 1st Qu. Median Mean 3rd Qu. Max. 0.00 9.00 14.67 16.37 22.00 51.00

TADPOLECrossMRI$ADAS11 <- NULL
TADPOLECrossMRI$MMSE <- NULL
TADPOLECrossMRI$RAVLT_immediate <- NULL
TADPOLECrossMRI$RAVLT_learning <- NULL
TADPOLECrossMRI$RAVLT_perc_forgetting <- NULL
TADPOLECrossMRI$RAVLT_forgetting <- NULL
TADPOLECrossMRI$FAQ <- NULL
TADPOLECrossMRI$DX <- NULL

trainSet <- sample(nrow(TADPOLECrossMRI),nrow(TADPOLECrossMRI)*TrainFraction)

TADPOLECrossMRITrain <- TADPOLECrossMRI[trainSet,]
TADPOLECrossMRITest <- TADPOLECrossMRI[-trainSet,]

2.1.1 Learning ADAS13

bml <- BSWiMS.model(ADAS13~.,TADPOLECrossMRITrain,maxTrainModelSize=50,NumberofRepeats = 20)

[+-++-+-++-++-+-+-+-++-++-++-++-++-+-+-++-+-+-++-++-]…

pander::pander(bml$bagging$Jaccard.SM)

0.307

fs <- bml$bagging$frequencyTable
barplot(fs[order(-fs)],las=2,main="Selected Features",cex.names = 0.5)

sm <- summary(bml)
pander::pander(sm$coefficients)

	Estimate	lower	mean	upper	u.MSE	r.MSE	model.MSE	NeRI	F.pvalue	t.pvalue	Sign.pvalue	Wilcox.pvalue	Frequency
M_ST24CV	-310.374	-354.175	-310.374	-266.5736	69.6	60.8	48.4	0.18453	0.00e+00	8.90e-17	5.24e-07	2.69e-09	0.55
M_ST24TA	-352.914	-447.773	-352.914	-258.0546	65.1	48.4	44.9	0.05832	1.53e-13	3.00e-05	5.62e-02	4.06e-03	1.00
M_ST24SA	85.592	61.655	85.592	109.5299	95.8	51.9	48.4	0.08729	1.21e-12	8.38e-06	9.44e-03	1.23e-03	0.55
M_ST51TA	829.670	578.616	829.670	1080.7246	92.7	48.6	45.8	0.05832	4.67e-11	5.82e-04	5.83e-02	2.30e-02	0.95
M_ST31TA	-583.062	-794.625	-583.062	-371.4987	80.8	49.1	47.0	0.06143	3.30e-08	1.28e-05	3.42e-02	3.67e-03	0.55
M_ST31CV	-80.805	-113.204	-80.805	-48.4049	78.5	46.5	44.9	0.04505	5.09e-07	2.69e-03	1.02e-01	4.26e-02	0.90
M_ST49TA	121.401	72.711	121.401	170.0912	92.7	49.1	47.4	0.06970	5.12e-07	1.18e-02	3.50e-02	5.37e-02	0.10
RD_ST32TA	15.084	8.937	15.084	21.2315	87.2	46.4	44.9	0.02333	7.57e-07	1.98e-02	2.66e-01	1.66e-01	1.00
Hippocampus	-107.597	-152.481	-107.597	-62.7136	69.6	46.3	44.9	0.11010	1.31e-06	4.17e-06	1.60e-03	4.48e-04	1.00
M_ST55CV	-100.224	-143.349	-100.224	-57.1001	85.1	48.5	47.1	0.03820	2.62e-06	9.38e-03	1.38e-01	5.56e-02	0.70
M_ST56TA	-84.415	-121.210	-84.415	-47.6190	86.7	48.0	46.7	0.01991	3.45e-06	3.14e-02	3.11e-01	1.69e-01	0.10
M_ST129TS	659.579	368.769	659.579	950.3892	95.7	49.0	47.6	0.05548	4.39e-06	3.67e-03	6.35e-02	3.53e-02	0.55
M_ST15TA	-337.926	-488.359	-337.926	-187.4931	85.4	46.7	45.4	0.05349	5.34e-06	1.90e-03	6.68e-02	3.71e-02	0.50
ST10CV	-0.154	-0.224	-0.154	-0.0842	97.5	46.9	45.6	0.07145	7.68e-06	4.43e-04	2.41e-02	1.32e-02	0.65
Gender	-0.813	-1.189	-0.813	-0.4370	96.6	48.7	47.5	0.07783	1.13e-05	1.44e-03	1.99e-02	1.64e-02	0.35
M_ST51CV	30.311	15.650	30.311	44.9730	94.0	49.0	47.8	0.05263	2.54e-05	5.54e-02	7.50e-02	8.52e-02	0.25
M_ST39SA	5.474	2.802	5.474	8.1453	96.7	46.5	45.5	0.03983	2.97e-05	5.62e-02	1.36e-01	9.92e-02	0.20
M_ST56TS	234.494	115.922	234.494	353.0649	96.1	48.6	47.6	0.06780	5.31e-05	3.85e-04	2.82e-02	1.11e-02	0.15
M_ST55SA	-12.775	-19.397	-12.775	-6.1540	94.9	46.6	45.6	0.04453	7.79e-05	1.14e-02	1.06e-01	8.72e-02	0.65
M_ST39CV	28.531	13.504	28.531	43.5576	92.9	47.6	46.7	0.02845	9.91e-05	6.86e-02	1.86e-01	1.29e-01	0.30
M_ST30SV	43.320	20.477	43.320	66.1625	73.3	46.0	45.1	0.06842	1.01e-04	8.17e-03	3.27e-02	2.34e-02	1.00
RD_ST35TA	6.856	3.215	6.856	10.4980	95.3	49.6	48.6	0.04315	1.12e-04	6.78e-02	1.27e-01	1.76e-01	0.30
RD_ST56TA	23.495	10.946	23.495	36.0441	94.6	46.1	45.2	0.03954	1.22e-04	7.62e-02	1.32e-01	1.71e-01	1.00
M_ST31SA	-6.608	-10.142	-6.608	-3.0745	92.4	48.6	47.7	0.01920	1.24e-04	1.13e-02	2.24e-01	9.99e-02	0.20
M_ST40CV	-58.917	-90.490	-58.917	-27.3432	72.7	49.3	48.3	0.01027	1.27e-04	1.94e-01	4.33e-01	2.74e-01	0.45
M_ST17SV	40.808	18.663	40.808	62.9521	97.4	46.1	45.3	0.06950	1.52e-04	2.79e-05	3.09e-02	2.31e-03	0.70
RD_ST31TA	13.935	6.314	13.935	21.5558	90.0	45.7	44.9	0.04083	1.69e-04	8.58e-03	1.35e-01	8.65e-02	1.00
RD_ST40CV	7.147	3.200	7.147	11.0945	92.6	46.6	45.8	0.10633	1.94e-04	7.16e-04	2.09e-03	4.01e-03	0.40
RD_ST57TA	16.858	7.530	16.858	26.1869	92.5	45.8	45.0	0.08065	1.99e-04	1.62e-02	1.67e-02	3.98e-02	1.00
RD_ST40TA	3.577	1.459	3.577	5.6958	88.3	49.1	48.4	0.04339	4.67e-04	1.17e-01	1.31e-01	1.69e-01	0.20
M_ST65SV	28.707	11.663	28.707	45.7502	95.8	46.0	45.3	0.05263	4.81e-04	4.17e-02	7.59e-02	8.55e-02	0.25
M_ST32CV	-22.114	-35.271	-22.114	-8.9575	73.3	46.8	46.1	0.02987	4.93e-04	5.20e-03	2.08e-01	5.79e-02	0.25
RD_ST26TA	5.896	2.355	5.896	9.4370	90.0	48.8	48.1	0.02906	5.50e-04	5.47e-03	2.11e-01	5.39e-02	0.35
M_ST23CV	24.683	9.497	24.683	39.8702	96.6	47.8	47.1	0.00284	7.22e-04	2.88e-01	5.66e-01	3.72e-01	0.20
Ventricles	-4.673	-7.626	-4.673	-1.7201	87.2	45.3	44.7	0.01849	9.62e-04	5.27e-02	2.92e-01	1.52e-01	0.30
M_ST34TA	-40.225	-66.960	-40.225	-13.4898	86.0	49.7	49.1	0.03414	1.59e-03	1.79e-02	1.93e-01	7.69e-02	0.10
WholeBrain	5.211	1.429	5.211	8.9934	83.3	48.7	48.2	0.03414	3.46e-03	1.01e-01	1.92e-01	1.77e-01	0.10
M_ST38TA	47.270	9.605	47.270	84.9354	91.6	46.9	46.4	-0.00427	6.95e-03	6.78e-01	6.18e-01	6.19e-01	0.10
M_ST43TS	156.038	30.546	156.038	281.5294	97.5	45.6	45.2	0.04523	7.40e-03	6.38e-02	1.14e-01	1.32e-01	0.25

pander::pander(bml$univariate[bml$selectedfeatures,])

	Name	RName	ZUni
M_ST24TA	M_ST24TA	M_ST24TA	Inf
RD_ST32TA	RD_ST32TA	RD_ST32TA	Inf
M_ST30SV	M_ST30SV	M_ST30SV	Inf
Ventricles	Ventricles	Ventricles	Inf
Hippocampus	Hippocampus	Hippocampus	Inf
M_ST18SV	M_ST18SV	M_ST18SV	2.11
RD_ST56TA	RD_ST56TA	RD_ST56TA	4.85
RD_ST57TA	RD_ST57TA	RD_ST57TA	6.27
RD_ST31TA	RD_ST31TA	RD_ST31TA	7.63
M_ST43TS	M_ST43TS	M_ST43TS	1.48
M_ST31CV	M_ST31CV	M_ST31CV	Inf
M_ST65SV	M_ST65SV	M_ST65SV	3.82
M_ST55SA	M_ST55SA	M_ST55SA	4.61
M_ST32CV	M_ST32CV	M_ST32CV	Inf
M_ST60CV	M_ST60CV	M_ST60CV	Inf
M_ST39SA	M_ST39SA	M_ST39SA	2.86
M_ST17SV	M_ST17SV	M_ST17SV	1.76
M_ST51TA	M_ST51TA	M_ST51TA	6.10
M_ST15TA	M_ST15TA	M_ST15TA	Inf
M_ST24CV	M_ST24CV	M_ST24CV	Inf
M_ST55CV	M_ST55CV	M_ST55CV	Inf
RD_ST40CV	RD_ST40CV	RD_ST40CV	6.17
WholeBrain	WholeBrain	WholeBrain	Inf
ST10CV	ST10CV	ST10CV	1.54
M_ST50TA	M_ST50TA	M_ST50TA	7.20
M_ST24SA	M_ST24SA	M_ST24SA	3.75
M_ST40CV	M_ST40CV	M_ST40CV	Inf
M_ST39CV	M_ST39CV	M_ST39CV	5.98
M_ST38TA	M_ST38TA	M_ST38TA	6.77
M_ST129TS	M_ST129TS	M_ST129TS	3.89
M_ST51CV	M_ST51CV	M_ST51CV	5.27
RD_ST26TA	RD_ST26TA	RD_ST26TA	7.62
M_ST23CV	M_ST23CV	M_ST23CV	2.91
M_ST56TS	M_ST56TS	M_ST56TS	3.45
M_ST56TA	M_ST56TA	M_ST56TA	Inf
RD_ST35TA	RD_ST35TA	RD_ST35TA	4.22
M_ST31TA	M_ST31TA	M_ST31TA	Inf
RD_ST32CV	RD_ST32CV	RD_ST32CV	5.50
RD_ST40TA	RD_ST40TA	RD_ST40TA	Inf
M_ST40TS	M_ST40TS	M_ST40TS	1.22
M_ST31SA	M_ST31SA	M_ST31SA	6.30
M_ST40TA	M_ST40TA	M_ST40TA	Inf
Gender	Gender	Gender	2.98
M_ST49TA	M_ST49TA	M_ST49TA	6.15
M_ST15CV	M_ST15CV	M_ST15CV	6.97
M_ST34TA	M_ST34TA	M_ST34TA	Inf
M_ST35SA	M_ST35SA	M_ST35SA	2.11
M_ST35CV	M_ST35CV	M_ST35CV	7.30

prreg <- predictionStats_regression(cbind(TADPOLECrossMRITest$ADAS13,predict(bml,TADPOLECrossMRITest)),"ADAS13")

ADAS13

pander::pander(prreg)

• corci:

  cor
  0.705	0.666	0.74

• biasci: 0.803, 0.330 and 1.276

• RMSEci: 6.44, 6.13 and 6.80

• spearmanci:

  50%	2.5%	97.5%
  0.627	0.576	0.673

• MAEci:

  50%	2.5%	97.5%
  5.13	4.84	5.42

• pearson:

  Pearson’s product-moment correlation: predictions[, 1] and predictions[, 2]

  Test statistic	df	P value	Alternative hypothesis	cor
  26.3	702	9.41e-107 * * *	two.sided	0.705

par(op)

2.1.2 The formula network

cmax <- apply(bml$bagging$formulaNetwork,2,max)
cnames <- names(cmax[cmax>=0.250])
cmax <- cmax[cnames]

adma <- bml$bagging$formulaNetwork[cnames,cnames]

adma[adma<0.15] <- 0;
gr <- graph_from_adjacency_matrix(adma,mode = "undirected",diag = FALSE,weighted=TRUE)
gr$layout <- layout_with_fr

fc <- cluster_optimal(gr)
plot(fc, gr,
     edge.width=5*E(gr)$weight,
     vertex.size=20*cmax,
     vertex.label.cex=0.5,
     vertex.label.dist=0,
     main="ADAS13 Feature Association")

par(op)

2.1.3 The ADAS13 table

clusterFeatures <- fc$names
sm$coefficients$Rx2 <- (sm$coefficients$r.MSE-sm$coefficients$model.MSE)/sm$coefficients$r.MSE

tableADAS13 <- sm$coefficients[clusterFeatures,
                                   c("Estimate",
                                     "lower",
                                     "mean",
                                     "upper",
                                     "model.MSE",
                                     "Rx2",
                                     "F.pvalue",
                                     "Frequency")]

nugget <- fc$membership
names(nugget) <- clusterFeatures

tableADAS13$Cluster <- nugget[rownames(tableADAS13)]

rnames <- clusterFeatures[str_detect(clusterFeatures,"ST")]
frnames <- rnames
rnames <- str_replace_all(rnames,"M_","")
rnames <- str_replace_all(rnames,"RD_","")
description <- character()

for (ddet in c(1:length(rnames)))
{
  description <- c(description,TADPOLE_D1_D2_Dict$TEXT[str_detect(TADPOLE_D1_D2_Dict$FLDNAME,rnames[ddet])][1])
}
names(description) <- frnames

tableADAS13$Description <- description[rownames(tableADAS13)]
pander::pander(tableADAS13)

	Estimate	lower	mean	upper	model.MSE	Rx2	F.pvalue	Frequency	Cluster	Description
M_ST24TA	-352.914	-447.773	-352.914	-258.0546	44.9	0.07294	1.53e-13	1.00	1	Cortical Thickness Average of LeftEntorhinal
RD_ST32TA	15.084	8.937	15.084	21.2315	44.9	0.03246	7.57e-07	1.00	1	Cortical Thickness Average of LeftInferiorTemporal
M_ST30SV	43.320	20.477	43.320	66.1625	45.1	0.01966	1.01e-04	1.00	1	Volume (WM Parcellation) of LeftInferiorLateralVentricle
Ventricles	-4.673	-7.626	-4.673	-1.7201	44.7	0.01361	9.62e-04	0.30	1	NA
Hippocampus	-107.597	-152.481	-107.597	-62.7136	44.9	0.03097	1.31e-06	1.00	1	NA
RD_ST56TA	23.495	10.946	23.495	36.0441	45.2	0.01907	1.22e-04	1.00	1	Cortical Thickness Average of LeftSuperiorFrontal
RD_ST57TA	16.858	7.530	16.858	26.1869	45.0	0.01773	1.99e-04	1.00	1	Cortical Thickness Average of LeftSuperiorParietal
RD_ST31TA	13.935	6.314	13.935	21.5558	44.9	0.01815	1.69e-04	1.00	1	Cortical Thickness Average of LeftInferiorParietal
M_ST43TS	156.038	30.546	156.038	281.5294	45.2	0.00848	7.40e-03	0.25	1	Cortical Thickness Standard Deviation of LeftParacentral
M_ST31CV	-80.805	-113.204	-80.805	-48.4049	44.9	0.03371	5.09e-07	0.90	1	Volume (Cortical Parcellation) of LeftInferiorParietal
M_ST65SV	28.707	11.663	28.707	45.7502	45.3	0.01557	4.81e-04	0.25	1	Volume (WM Parcellation) of LeftVentralDC
M_ST55SA	-12.775	-19.397	-12.775	-6.1540	45.6	0.02029	7.79e-05	0.65	1	Surface Area of LeftRostralMiddleFrontal
M_ST32CV	-22.114	-35.271	-22.114	-8.9575	46.1	0.01552	4.93e-04	0.25	2	Volume (Cortical Parcellation) of LeftInferiorTemporal
M_ST17SV	40.808	18.663	40.808	62.9521	45.3	0.01845	1.52e-04	0.70	1	Volume (WM Parcellation) of LeftCerebellumCortex
M_ST51TA	829.670	578.616	829.670	1080.7246	45.8	0.05798	4.67e-11	0.95	2	Cortical Thickness Average of LeftPrecentral
M_ST15TA	-337.926	-488.359	-337.926	-187.4931	45.4	0.02721	5.34e-06	0.50	1	Cortical Thickness Average of LeftCaudalMiddleFrontal
M_ST24CV	-310.374	-354.175	-310.374	-266.5736	48.4	0.20433	0.00e+00	0.55	2	Volume (Cortical Parcellation) of LeftEntorhinal
M_ST55CV	-100.224	-143.349	-100.224	-57.1001	47.1	0.02943	2.62e-06	0.70	2	Volume (Cortical Parcellation) of LeftRostralMiddleFrontal
RD_ST40CV	7.147	3.200	7.147	11.0945	45.8	0.01782	1.94e-04	0.40	2	Volume (Cortical Parcellation) of LeftMiddleTemporal
ST10CV	-0.154	-0.224	-0.154	-0.0842	45.6	0.02665	7.68e-06	0.65	2	Volume (Cortical Parcellation) of Icv
M_ST24SA	85.592	61.655	85.592	109.5299	48.4	0.06746	1.21e-12	0.55	2	Surface Area of LeftEntorhinal
M_ST40CV	-58.917	-90.490	-58.917	-27.3432	48.3	0.01922	1.27e-04	0.45	2	Volume (Cortical Parcellation) of LeftMiddleTemporal
M_ST39CV	28.531	13.504	28.531	43.5576	46.7	0.01954	9.91e-05	0.30	1	Volume (Cortical Parcellation) of LeftMedialOrbitofrontal
M_ST129TS	659.579	368.769	659.579	950.3892	47.6	0.02833	4.39e-06	0.55	2	Cortical Thickness Standard Deviation of LeftInsula
M_ST51CV	30.311	15.650	30.311	44.9730	47.8	0.02324	2.54e-05	0.25	2	Volume (Cortical Parcellation) of LeftPrecentral
RD_ST26TA	5.896	2.355	5.896	9.4370	48.1	0.01513	5.50e-04	0.35	2	Cortical Thickness Average of LeftFusiform
RD_ST35TA	6.856	3.215	6.856	10.4980	48.6	0.01919	1.12e-04	0.30	2	Cortical Thickness Average of LeftLateralOccipital
M_ST31TA	-583.062	-794.625	-583.062	-371.4987	47.0	0.04105	3.30e-08	0.55	2	Cortical Thickness Average of LeftInferiorParietal
Gender	-0.813	-1.189	-0.813	-0.4370	47.5	0.02538	1.13e-05	0.35	2	NA

2.1.4 Decorrelating training and testing sets

TADPOLECrossMRITrainD <- GDSTMDecorrelation(TADPOLECrossMRITrain,Outcome="ADAS13",
                                        thr=0.6,
                                        type="RLM",
                                        method="spearman",
                                        verbose = TRUE)
#> 
#>  Included: 197 , Uni p: 0.01145966 To Outcome: 122 , Base: 6 , In Included: 6 , Base Cor: 18 
#> 1 , Top: 36 < 0.6 >( 2 )[ 1 : 0 : 0.594 ]( 35 , 88 , 0 ),<>Tot Used: 123 , Added: 88 , Zero Std: 0 , Max Cor: 0.9484249 
#> 2 , Top: 25 < 0.6 >( 1 )[ 1 : 0 : 0.594 ]( 24 , 35 , 35 ),<>Tot Used: 147 , Added: 35 , Zero Std: 0 , Max Cor: 0.8993252 
#> 3 , Top: 19 < 0.6 >( 2 )[ 1 : 0 : 0 ]( 16 , 18 , 54 ),<>Tot Used: 151 , Added: 18 , Zero Std: 0 , Max Cor: 0.8960537 
#> 4 , Top: 5 < 0.6 >[ TRUE ]( 1 )[ 1 : 0 : 0 ]( 5 , 5 , 65 ),<>Tot Used: 151 , Added: 5 , Zero Std: 0 , Max Cor: 0.6139909 
#> 5 , Top: 2 < 0.6 >[ FALSE ]( 1 )[ 1 : 0 : 0.6 ]( 2 , 2 , 67 ),<>Tot Used: 151 , Added: 2 , Zero Std: 0 , Max Cor: 0.6288532 
#> 6 , Top: 1 < 0.6 >( 1 )[ 1 : 0 : 0.6 ]( 1 , 1 , 68 ),<>Tot Used: 151 , Added: 1 , Zero Std: 0 , Max Cor: 0.5981779 
#> [ 7 ], 0.5981779 . Cor to Base: 97 , ABase: 68
TADPOLECrossMRITestD <-  predictDecorrelate(TADPOLECrossMRITrainD,TADPOLECrossMRITest)

2.2 Decorrelated

bmlD <- BSWiMS.model(ADAS13~.,TADPOLECrossMRITrainD,maxTrainModelSize=50,NumberofRepeats = 20)

[+-+–+-+-+-+-++-+-+-+-+-+-+-+-+-++-+-+-+-+-]..

pander::pander(bmlD$bagging$Jaccard.SM)

0.549

fs <- bmlD$bagging$frequencyTable
barplot(fs[order(-fs)],las=2,main="Selected Features",cex.names = 0.5)

sm <- summary(bmlD)
pander::pander(sm$coefficients)

	Estimate	lower	mean	upper	u.MSE	r.MSE	model.MSE	NeRI	F.pvalue	t.pvalue	Sign.pvalue	Wilcox.pvalue	Frequency
Ba_Hippocampus	-415.962	-4.75e+02	-415.962	-357.2610	69.6	60.9	45.0	0.22248	0.00e+00	1.78e-26	1.78e-09	9.21e-14	1.00
De_M_ST24TA	-699.354	-8.04e+02	-699.354	-594.8917	90.2	50.0	45.0	0.06785	0.00e+00	1.51e-05	3.44e-02	2.61e-03	1.00
Ba_M_ST59TA	-291.416	-3.41e+02	-291.416	-241.7209	83.6	55.4	48.1	0.13324	0.00e+00	4.04e-15	2.31e-05	1.00e-07	0.15
De_M_ST32TA	-136.544	-1.75e+02	-136.544	-97.9165	90.8	53.8	50.3	0.09104	2.13e-12	1.74e-05	7.81e-03	7.42e-04	0.10
De_M_ST26TA	173.749	1.22e+02	173.749	225.4056	96.2	53.5	50.3	0.13514	2.16e-11	4.17e-06	1.88e-04	1.15e-04	0.10
Ba_ICV	-0.135	-1.79e-01	-0.135	-0.0906	97.5	48.7	46.3	0.11901	1.24e-09	1.98e-07	8.54e-04	1.81e-04	0.30
De_M_ST26CV	-43.738	-6.01e+01	-43.738	-27.4099	95.3	52.3	50.3	0.08250	7.59e-08	5.64e-05	1.55e-02	2.44e-03	0.10
De_M_ST24CV	-20.033	-2.77e+01	-20.033	-12.3308	92.1	52.2	50.3	0.05548	1.72e-07	5.73e-03	7.57e-02	4.03e-02	0.10
RD_ST32TA	26.312	1.57e+01	26.312	36.9143	87.2	46.6	45.0	0.01494	5.75e-07	4.02e-02	3.37e-01	2.60e-01	1.00
Ba_RD_ST40CV	3.127	1.85e+00	3.127	4.4092	92.6	51.9	50.3	0.08962	8.69e-07	1.26e-03	8.18e-03	3.68e-03	0.10
De_M_ST51TA	1205.483	6.86e+02	1205.483	1724.7535	97.0	46.4	45.0	0.03442	2.68e-06	2.36e-02	1.76e-01	1.07e-01	1.00
De_M_ST31TA	-1161.218	-1.70e+03	-1161.218	-626.1671	94.6	46.2	45.1	0.05184	1.05e-05	3.41e-03	8.40e-02	2.93e-02	0.90
De_Ventricles	291.033	1.52e+02	291.033	430.0559	89.0	46.1	45.0	0.05960	2.04e-05	2.78e-03	4.92e-02	1.17e-02	1.00
De_M_ST43TA	119.652	6.25e+01	119.652	176.8476	96.5	45.5	44.4	0.05358	2.06e-05	7.93e-03	8.10e-02	3.96e-02	0.15
RD_ST56TA	41.539	2.13e+01	41.539	61.7382	94.6	46.1	45.0	0.03570	2.78e-05	4.05e-02	1.66e-01	1.31e-01	1.00
De_M_ST31SA	-24.610	-3.69e+01	-24.610	-12.3319	92.4	46.6	45.6	0.04315	4.27e-05	7.48e-03	1.28e-01	6.52e-02	0.75
RD_ST31TA	23.748	1.19e+01	23.748	35.6319	90.0	46.1	45.1	0.03035	4.49e-05	6.11e-03	2.09e-01	7.98e-02	0.90
Gender	-0.667	-1.00e+00	-0.667	-0.3311	96.6	47.3	46.3	0.07207	5.02e-05	1.90e-04	2.42e-02	5.62e-03	0.30
De_M_ST40TA	-70.203	-1.07e+02	-70.203	-33.4820	97.4	45.9	44.9	0.06401	8.95e-05	9.19e-02	3.97e-02	7.65e-02	0.10
RD_ST57TA	27.604	1.32e+01	27.604	42.0533	92.5	46.0	45.1	0.05862	9.04e-05	8.87e-02	5.84e-02	1.13e-01	0.95
De_M_ST55SA	-21.673	-3.31e+01	-21.673	-10.2440	94.9	46.0	45.1	0.03917	1.01e-04	3.24e-02	1.34e-01	1.60e-01	0.75
M_ST129TS	503.213	2.32e+02	503.213	774.0919	95.7	46.2	45.3	0.05832	1.36e-04	9.12e-04	6.23e-02	2.03e-02	0.45
De_M_ST38TA	463.619	2.06e+02	463.619	720.8942	97.6	46.0	45.2	0.01394	2.06e-04	9.95e-02	3.85e-01	2.10e-01	0.50
De_RD_ST49CV	4.484	1.97e+00	4.484	6.9948	97.0	44.9	44.1	-0.00569	2.32e-04	1.02e-01	4.99e-01	1.83e-01	0.10
M_ST17SV	25.152	1.10e+01	25.152	39.3426	97.4	45.8	45.0	0.11901	2.56e-04	3.36e-06	8.54e-04	6.03e-04	0.30
De_M_ST49TA	180.140	7.60e+01	180.140	284.3079	95.6	46.2	45.4	0.05334	3.50e-04	1.42e-01	8.23e-02	1.74e-01	0.20
RD_ST26TA	2.842	1.12e+00	2.842	4.5642	90.0	49.0	48.3	0.00711	6.10e-04	9.23e-02	4.84e-01	2.02e-01	0.15
De_M_ST58TA	64.756	2.21e+01	64.756	107.4144	94.6	50.9	50.3	0.00427	1.46e-03	5.74e-02	4.70e-01	2.61e-01	0.10
Ba_M_ST21SV	11.668	3.93e+00	11.668	19.4031	87.7	50.9	50.3	0.03129	1.55e-03	7.63e-02	2.05e-01	1.73e-01	0.10
M_ST43TS	501.764	1.53e+02	501.764	850.9519	97.5	45.5	45.0	0.02347	2.43e-03	4.79e-02	2.54e-01	1.52e-01	0.40
De_RD_ST31CV	7.651	1.57e+00	7.651	13.7292	95.6	45.8	45.4	-0.01328	6.81e-03	8.93e-02	5.86e-01	3.16e-01	0.30
RD_ST24TA	0.611	6.04e-02	0.611	1.1624	88.8	50.6	50.3	0.04125	1.48e-02	4.13e-01	1.42e-01	3.34e-01	0.10

pander::pander(bmlD$univariate[bmlD$selectedfeatures,])

	Name	RName	ZUni
Ba_Hippocampus	Ba_Hippocampus	Ba_Hippocampus	Inf
De_M_ST24TA	De_M_ST24TA	De_M_ST24TA	7.52
De_Ventricles	De_Ventricles	De_Ventricles	8.13
RD_ST32TA	RD_ST32TA	RD_ST32TA	Inf
De_M_ST55SA	De_M_ST55SA	De_M_ST55SA	4.61
De_RD_ST31CV	De_RD_ST31CV	De_RD_ST31CV	3.99
De_M_ST51TA	De_M_ST51TA	De_M_ST51TA	2.38
RD_ST57TA	RD_ST57TA	RD_ST57TA	6.27
De_M_ST31TA	De_M_ST31TA	De_M_ST31TA	4.79
RD_ST56TA	RD_ST56TA	RD_ST56TA	4.85
RD_ST31TA	RD_ST31TA	RD_ST31TA	7.63
M_ST43TS	M_ST43TS	M_ST43TS	1.48
De_M_ST31SA	De_M_ST31SA	De_M_ST31SA	6.30
M_ST129TS	M_ST129TS	M_ST129TS	3.89
De_M_ST38TA	De_M_ST38TA	De_M_ST38TA	1.19
De_M_ST23TA	De_M_ST23TA	De_M_ST23TA	3.80
M_ST60SA	M_ST60SA	M_ST60SA	2.76
De_RD_ST49CV	De_RD_ST49CV	De_RD_ST49CV	2.39
M_ST17SV	M_ST17SV	M_ST17SV	1.76
De_M_ST43TA	De_M_ST43TA	De_M_ST43TA	2.99
De_M_ST40TA	De_M_ST40TA	De_M_ST40TA	1.58
De_M_ST32TA	De_M_ST32TA	De_M_ST32TA	7.19
Ba_M_ST59TA	Ba_M_ST59TA	Ba_M_ST59TA	Inf
Ba_ICV	Ba_ICV	Ba_ICV	1.53
RD_ST24TA	RD_ST24TA	RD_ST24TA	8.21
De_M_ST24CV	De_M_ST24CV	De_M_ST24CV	6.51
RD_ST26TA	RD_ST26TA	RD_ST26TA	7.62
De_M_ST49TA	De_M_ST49TA	De_M_ST49TA	3.95
Gender	Gender	Gender	2.98
Ba_M_ST21SV	Ba_M_ST21SV	Ba_M_ST21SV	Inf
Ba_RD_ST32CV	Ba_RD_ST32CV	Ba_RD_ST32CV	5.50
De_M_ST40SA	De_M_ST40SA	De_M_ST40SA	7.46
De_M_ST58TA	De_M_ST58TA	De_M_ST58TA	4.83
De_M_ST26TA	De_M_ST26TA	De_M_ST26TA	3.39
Ba_RD_ST40CV	Ba_RD_ST40CV	Ba_RD_ST40CV	6.17
De_M_ST26CV	De_M_ST26CV	De_M_ST26CV	4.22
De_M_ST51CV	De_M_ST51CV	De_M_ST51CV	2.20
Ba_M_ST24SA	Ba_M_ST24SA	Ba_M_ST24SA	3.75
M_ST47TS	M_ST47TS	M_ST47TS	1.18
Ba_M_ST35SA	Ba_M_ST35SA	Ba_M_ST35SA	2.11

prreg <- predictionStats_regression(cbind(TADPOLECrossMRITestD$ADAS13,predict(bmlD,TADPOLECrossMRITestD)),"ADAS13")

ADAS13

pander::pander(prreg)

• corci:

  cor
  0.678	0.636	0.716

• biasci: 0.956, 0.460 and 1.453

• RMSEci: 6.78, 6.44 and 7.15

• spearmanci:

  50%	2.5%	97.5%
  0.614	0.561	0.661

• MAEci:

  50%	2.5%	97.5%
  5.35	5.06	5.67

• pearson:

  Pearson’s product-moment correlation: predictions[, 1] and predictions[, 2]

  Test statistic	df	P value	Alternative hypothesis	cor
  24.4	702	8.7e-96 * * *	two.sided	0.678

par(op)

2.2.1 The formula network

cmax <- apply(bmlD$bagging$formulaNetwork,2,max)
cnames <- names(cmax[cmax>=0.250])
cmax <- cmax[cnames]

adma <- bmlD$bagging$formulaNetwork[cnames,cnames]

adma[adma<0.15] <- 0;
gr <- graph_from_adjacency_matrix(adma,mode = "undirected",diag = FALSE,weighted=TRUE)
gr$layout <- layout_with_fr

fc <- cluster_optimal(gr)
plot(fc, gr,
     edge.width=5*E(gr)$weight,
     vertex.size=20*cmax,
     vertex.label.cex=0.5,
     vertex.label.dist=0,
     main="ADAS13 Feature Association")

par(op)

2.2.2 Decorrelated ADAS13 table

clusterFeatures <- fc$names
sm$coefficients$Rx2 <- (sm$coefficients$r.MSE-sm$coefficients$model.MSE)/sm$coefficients$r.MSE

tableADAS13D <- sm$coefficients[clusterFeatures,
                                   c("Estimate",
                                     "lower",
                                     "mean",
                                     "upper",
                                     "model.MSE",
                                     "Rx2",
                                     "F.pvalue",
                                     "Frequency")]

nugget <- fc$membership
names(nugget) <- clusterFeatures

tableADAS13D$Cluster <- nugget[rownames(tableADAS13D)]

rnames <- clusterFeatures[str_detect(clusterFeatures,"ST")]
frnames <- rnames
rnames <- str_replace_all(rnames,"M_","")
rnames <- str_replace_all(rnames,"RD_","")
rnames <- str_replace_all(rnames,"Ba_","")
rnames <- str_replace_all(rnames,"De_","")
description <- character()

for (ddet in c(1:length(rnames)))
{
  description <- c(description,TADPOLE_D1_D2_Dict$TEXT[str_detect(TADPOLE_D1_D2_Dict$FLDNAME,rnames[ddet])][1])
}
names(description) <- frnames

tableADAS13D$Description <- description[rownames(tableADAS13D)]

## Getting the decorrelation formula
dc <- getDerivedCoefficients(TADPOLECrossMRITrainD)
decornames <- rownames(sm$coefficients)

deNames_in_dc <- decornames[decornames %in% names(dc)]
theDeFormulas <- dc[deNames_in_dc]
deFromula <- character(length(theDeFormulas))
names(deFromula) <- names(theDeFormulas)
for (dx in names(deFromula))
{
  coef <- theDeFormulas[[dx]]
  cname <- names(theDeFormulas[[dx]])
  names(cname) <- cname
  for (cf in names(coef))
  {
    if (cf != dx)
    {
      if (coef[cf]>0)
      {
        deFromula[dx] <- paste(deFromula[dx],
                               sprintf("+ %5.3f*%s",coef[cf],cname[cf]))
      }
      else
      {
        deFromula[dx] <- paste(deFromula[dx],
                               sprintf("%5.3f*%s",coef[cf],cname[cf]))
      }
    }
  }
}

tableADAS13D$DecorFormula <- deFromula[rownames(tableADAS13D)]


pander::pander(tableADAS13D)

	Estimate	lower	mean	upper	model.MSE	Rx2	F.pvalue	Frequency	Cluster	Description	DecorFormula
Ba_Hippocampus	-415.962	-474.664	-415.962	-357.2610	45.0	0.26090	0.00e+00	1.00	1	NA	NA
De_M_ST24TA	-699.354	-803.816	-699.354	-594.8917	45.0	0.09919	0.00e+00	1.00	1	Cortical Thickness Average of LeftEntorhinal	-0.306Hippocampus + 1.000M_ST24TA
De_Ventricles	291.033	152.010	291.033	430.0559	45.0	0.02374	2.04e-05	1.00	1	NA	+ 1.000Ventricles -1.284M_ST37SV
RD_ST32TA	26.312	15.709	26.312	36.9143	45.0	0.03317	5.75e-07	1.00	1	Cortical Thickness Average of LeftInferiorTemporal	NA
De_M_ST55SA	-21.673	-33.102	-21.673	-10.2440	45.1	0.01958	1.01e-04	0.75	1	Surface Area of LeftRostralMiddleFrontal	NA
De_RD_ST31CV	7.651	1.573	7.651	13.7292	45.4	0.00866	6.81e-03	0.30	2	Volume (Cortical Parcellation) of LeftInferiorParietal	-0.644RD_ST31SA + 1.000RD_ST31CV
De_M_ST51TA	1205.483	686.212	1205.483	1724.7535	45.0	0.02911	2.68e-06	1.00	1	Cortical Thickness Average of LeftPrecentral	-0.676M_ST43TA + 1.000M_ST51TA -0.427*M_ST59TA
RD_ST57TA	27.604	13.155	27.604	42.0533	45.1	0.01977	9.04e-05	0.95	1	Cortical Thickness Average of LeftSuperiorParietal	NA
De_M_ST31TA	-1161.218	-1696.268	-1161.218	-626.1671	45.1	0.02542	1.05e-05	0.90	1	Cortical Thickness Average of LeftInferiorParietal	+ 1.000M_ST31TA -0.918M_ST59TA
RD_ST56TA	41.539	21.340	41.539	61.7382	45.0	0.02286	2.78e-05	1.00	1	Cortical Thickness Average of LeftSuperiorFrontal	NA
RD_ST31TA	23.748	11.864	23.748	35.6319	45.1	0.02163	4.49e-05	0.90	1	Cortical Thickness Average of LeftInferiorParietal	NA
M_ST43TS	501.764	152.575	501.764	850.9519	45.0	0.01126	2.43e-03	0.40	2	Cortical Thickness Standard Deviation of LeftParacentral	NA
De_M_ST31SA	-24.610	-36.887	-24.610	-12.3319	45.6	0.02211	4.27e-05	0.75	2	Surface Area of LeftInferiorParietal	NA
M_ST129TS	503.213	232.334	503.213	774.0919	45.3	0.01872	1.36e-04	0.45	1	Cortical Thickness Standard Deviation of LeftInsula	NA
De_M_ST38TA	463.619	206.344	463.619	720.8942	45.2	0.01762	2.06e-04	0.50	1	Cortical Thickness Average of LeftLingual	+ 1.000M_ST38TA -0.576M_ST59TA
M_ST17SV	25.152	10.961	25.152	39.3426	45.0	0.01704	2.56e-04	0.30	1	Volume (WM Parcellation) of LeftCerebellumCortex	NA
Ba_ICV	-0.135	-0.179	-0.135	-0.0906	46.3	0.04999	1.24e-09	0.30	2	NA	NA
Gender	-0.667	-1.004	-0.667	-0.3311	46.3	0.02132	5.02e-05	0.30	2	NA	NA

3 Diagnosis MCI vs AD

3.0.1 the set

TrainFraction <- 0.60;

TADPOLECrossMRI <- subset(validBaselineTadpole,DX == "Dementia" | DX == "MCI")
table(TADPOLECrossMRI$DX)

Dementia MCI NL 244 711 0

TADPOLECrossMRI$DX <- 1*(as.character(TADPOLECrossMRI$DX) == "Dementia")
table(TADPOLECrossMRI$DX)

0 1 711 244

TADPOLECrossMRI$ADAS13 <- NULL
TADPOLECrossMRI$ADAS11 <- NULL
TADPOLECrossMRI$MMSE <- NULL
TADPOLECrossMRI$RAVLT_immediate <- NULL
TADPOLECrossMRI$RAVLT_learning <- NULL
TADPOLECrossMRI$RAVLT_perc_forgetting <- NULL
TADPOLECrossMRI$RAVLT_forgetting <- NULL
TADPOLECrossMRI$FAQ <- NULL

TADPOLE_Cases <- subset(TADPOLECrossMRI,DX==1)
TADPOLE_Controls <- subset(TADPOLECrossMRI,DX==0)
trainCasesSet <- sample(nrow(TADPOLE_Cases),nrow(TADPOLE_Cases)*TrainFraction)
trainControlSet <- sample(nrow(TADPOLE_Controls),nrow(TADPOLE_Controls)*TrainFraction)

TADPOLE_DX_TRAIN <- rbind(TADPOLE_Cases[trainCasesSet,],TADPOLE_Controls[trainControlSet,])
TADPOLE_DX_TEST <- TADPOLECrossMRI[!(rownames(TADPOLECrossMRI) %in% rownames(TADPOLE_DX_TRAIN)),]
table(TADPOLE_DX_TEST$DX)

0 1 285 98

par(op)

3.0.2 Learning

bDXml <- BSWiMS.model(DX~.,TADPOLE_DX_TRAIN,NumberofRepeats = 20)

[+++-++++++—+++++-+++-+++-++-+++–+-++++-++++-+-++++–+++-++++-+–+++++–+++–+++-++++-++++-++++-+++++–+++-+++–]……..

pander::pander(bDXml$bagging$Jaccard.SM)

0.123

fs <- bDXml$bagging$frequencyTable
barplot(fs[order(-fs)],las=2,main="Selected Features",cex.names = 0.5)

sm <- summary(bDXml)
pander::pander(sm$coefficients)

	Estimate	lower	OR	upper	u.Accuracy	r.Accuracy	full.Accuracy	u.AUC	r.AUC	full.AUC	IDI	NRI	z.IDI	z.NRI	Frequency
M_ST24TA	-72.1718	1.71e-36	4.53e-32	1.20e-27	0.721	0.663	0.744	0.722	0.652	0.749	0.1677	0.828	13.34	13.36	1.00
M_ST40TA	-55.4204	1.29e-28	8.54e-25	5.64e-21	0.678	0.666	0.723	0.671	0.663	0.719	0.0922	0.588	9.38	9.04	0.45
M_ST24CV	-27.2158	1.77e-14	1.51e-12	1.30e-10	0.714	0.699	0.745	0.707	0.692	0.745	0.0899	0.635	9.14	9.91	1.00
M_ST31TA	-28.9532	9.54e-16	2.67e-13	7.44e-11	0.655	0.736	0.742	0.645	0.727	0.746	0.0711	0.499	8.07	7.59	0.20
M_ST12SV	-25.8479	4.60e-14	5.95e-12	7.70e-10	0.692	0.701	0.736	0.701	0.696	0.739	0.0672	0.555	7.86	8.49	1.00
M_ST31CV	-20.2570	2.30e-11	1.59e-09	1.11e-07	0.662	0.724	0.742	0.654	0.718	0.747	0.0606	0.448	7.54	6.75	0.95
Hippocampus	-9.9733	5.00e-06	4.66e-05	4.35e-04	0.675	0.699	0.716	0.677	0.696	0.721	0.0606	0.535	7.32	8.14	0.65
M_ST52CV	-5.5476	1.14e-03	3.90e-03	1.33e-02	0.638	0.699	0.725	0.631	0.698	0.731	0.0589	0.421	7.12	6.35	0.20
M_ST32TA	-48.6612	7.79e-27	7.36e-22	6.95e-17	0.672	0.700	0.729	0.671	0.698	0.730	0.0578	0.468	6.97	7.09	0.70
M_ST30SV	2.1112	4.98e+00	8.26e+00	1.37e+01	0.648	0.689	0.712	0.645	0.682	0.713	0.0505	0.444	6.62	6.68	0.25
M_ST32CV	-13.7800	2.15e-08	1.04e-06	4.99e-05	0.678	0.712	0.737	0.677	0.713	0.738	0.0500	0.429	6.44	6.44	0.95
M_ST40CV	-7.7090	5.05e-05	4.49e-04	3.99e-03	0.696	0.714	0.728	0.703	0.717	0.733	0.0465	0.454	6.27	6.84	0.50
M_ST26CV	-1.6643	1.26e-01	1.89e-01	2.85e-01	0.664	0.696	0.715	0.656	0.698	0.718	0.0452	0.421	5.97	6.35	0.10
ICV	-0.0337	9.56e-01	9.67e-01	9.78e-01	0.514	0.720	0.739	0.526	0.728	0.743	0.0356	0.343	5.78	5.11	1.00
M_ST32SA	-1.0856	2.38e-01	3.38e-01	4.78e-01	0.604	0.691	0.720	0.595	0.688	0.712	0.0361	0.345	5.65	5.13	0.25
M_ST36TA	7.3869	1.20e+02	1.61e+03	2.17e+04	0.585	0.694	0.712	0.593	0.697	0.716	0.0362	0.400	5.51	5.98	0.10
M_ST39CV	14.8041	1.23e+04	2.69e+06	5.86e+08	0.589	0.715	0.736	0.588	0.722	0.740	0.0282	0.262	5.12	3.89	0.95
RD_ST52TA	1.5455	2.62e+00	4.69e+00	8.40e+00	0.608	0.727	0.743	0.559	0.729	0.742	0.0315	0.267	5.10	4.03	0.45
M_ST45TA	22.3568	7.28e+05	5.12e+09	3.60e+13	0.592	0.699	0.723	0.575	0.691	0.716	0.0267	0.349	4.87	5.19	0.25
M_ST25TA	7.0126	6.69e+01	1.11e+03	1.84e+04	0.533	0.711	0.717	0.522	0.707	0.719	0.0242	0.333	4.82	4.94	0.15
M_ST43TS	22.2526	5.20e+05	4.62e+09	4.10e+13	0.532	0.730	0.747	0.516	0.729	0.749	0.0261	0.191	4.72	2.82	0.10
M_ST46TS	116.2719	7.02e+29	3.14e+50	1.40e+71	0.562	0.725	0.740	0.546	0.725	0.741	0.0241	0.306	4.69	4.55	1.00
RD_ST32TA	0.4213	1.28e+00	1.52e+00	1.81e+00	0.642	0.705	0.727	0.607	0.711	0.731	0.0238	0.329	4.69	4.95	0.30
M_ST24SA	2.2443	3.49e+00	9.43e+00	2.55e+01	0.538	0.728	0.743	0.554	0.729	0.742	0.0200	0.222	4.34	3.26	0.25
M_ST39SA	3.0038	5.29e+00	2.02e+01	7.69e+01	0.536	0.724	0.742	0.537	0.728	0.744	0.0186	0.210	4.30	3.09	0.70
Ventricles	-0.1639	7.83e-01	8.49e-01	9.20e-01	0.600	0.693	0.706	0.605	0.699	0.712	0.0173	0.252	3.93	3.72	0.10
M_ST60SA	1.0763	1.73e+00	2.93e+00	4.97e+00	0.540	0.715	0.730	0.539	0.715	0.730	0.0158	0.200	3.89	2.95	0.20

pander::pander(bDXml$univariate[bDXml$selectedfeatures,])

	Name	RName	ZUni
M_ST24TA	M_ST24TA	M_ST24TA	10.41
M_ST31CV	M_ST31CV	M_ST31CV	7.60
M_ST24CV	M_ST24CV	M_ST24CV	9.39
RD_ST52TA	RD_ST52TA	RD_ST52TA	3.90
M_ST31TA	M_ST31TA	M_ST31TA	7.28
ICV	ICV	ICV	1.14
M_ST24SA	M_ST24SA	M_ST24SA	1.88
M_ST43TS	M_ST43TS	M_ST43TS	2.26
M_ST12SV	M_ST12SV	M_ST12SV	9.19
M_ST39CV	M_ST39CV	M_ST39CV	3.00
M_ST32CV	M_ST32CV	M_ST32CV	8.51
M_ST46TS	M_ST46TS	M_ST46TS	2.05
M_ST40CV	M_ST40CV	M_ST40CV	8.30
M_ST60SA	M_ST60SA	M_ST60SA	1.64
M_ST39SA	M_ST39SA	M_ST39SA	1.46
M_ST32TA	M_ST32TA	M_ST32TA	8.31
RD_ST32TA	RD_ST32TA	RD_ST32TA	4.19
Hippocampus	Hippocampus	Hippocampus	8.55
M_ST52CV	M_ST52CV	M_ST52CV	7.26
M_ST40TS	M_ST40TS	M_ST40TS	1.61
M_ST36TA	M_ST36TA	M_ST36TA	3.47
M_ST26CV	M_ST26CV	M_ST26CV	7.77
M_ST40TA	M_ST40TA	M_ST40TA	7.71
M_ST30SV	M_ST30SV	M_ST30SV	7.61
M_ST32SA	M_ST32SA	M_ST32SA	5.06
M_ST25TA	M_ST25TA	M_ST25TA	1.24
M_ST45TA	M_ST45TA	M_ST45TA	4.02
Ventricles	Ventricles	Ventricles	4.29

prBin <- predictionStats_binary(cbind(TADPOLE_DX_TEST$DX,predict(bDXml,TADPOLE_DX_TEST)),"MCI vs Dementia")

MCI vs Dementia

pander::pander(prBin$aucs)

est	lower	upper
0.816	0.77	0.862

pander::pander(prBin$accc)

est	lower	upper
0.749	0.703	0.792

pander::pander(prBin$berror)

50%	2.5%	97.5%
0.238	0.194	0.288

pander::pander(prBin$sensitivity)

est	lower	upper
0.786	0.691	0.862

par(op)

3.0.3 The formula network

cmax <- apply(bDXml$bagging$formulaNetwork,2,max)
cnames <- names(cmax[cmax>=0.25])
cmax <- cmax[cnames]

adma <- bDXml$bagging$formulaNetwork[cnames,cnames]

adma[adma<0.15] <- 0;
gr <- graph_from_adjacency_matrix(adma,mode = "undirected",diag = FALSE,weighted=TRUE)
gr$layout <- layout_with_fr

fc <- cluster_optimal(gr)
plot(fc, gr,
     edge.width=10*E(gr)$weight,
     vertex.size=10*cmax,
     vertex.label.cex=0.75,
     vertex.label.dist=0,
     main="MCI vs Dementia Diagnosis")

par(op)

3.0.4 MCI vs AD table

clusterFeatures <- fc$names
sm$coefficients$DeltaAUC <- (sm$coefficients$full.AUC-sm$coefficients$r.AUC)

tableMCI_De <- sm$coefficients[clusterFeatures,
                                   c("Estimate",
                                     "lower",
                                     "OR",
                                     "upper",
                                     "full.AUC",
                                     "DeltaAUC",
                                     "z.IDI",
                                     "Frequency")]

nugget <- fc$membership
names(nugget) <- clusterFeatures

tableMCI_De$Cluster <- nugget[rownames(tableMCI_De)]

rnames <- clusterFeatures[str_detect(clusterFeatures,"ST")]
frnames <- rnames
rnames <- str_replace_all(rnames,"M_","")
rnames <- str_replace_all(rnames,"RD_","")
description <- character()

for (ddet in c(1:length(rnames)))
{
  description <- c(description,TADPOLE_D1_D2_Dict$TEXT[str_detect(TADPOLE_D1_D2_Dict$FLDNAME,rnames[ddet])][1])
}
names(description) <- frnames

tableMCI_De$Description <- description[rownames(tableMCI_De)]
pander::pander(tableMCI_De)

	Estimate	lower	OR	upper	full.AUC	DeltaAUC	z.IDI	Frequency	Cluster	Description
M_ST24TA	-72.1718	1.71e-36	4.53e-32	1.20e-27	0.749	0.0970	13.34	1.00	1	Cortical Thickness Average of LeftEntorhinal
M_ST31CV	-20.2570	2.30e-11	1.59e-09	1.11e-07	0.747	0.0298	7.54	0.95	1	Volume (Cortical Parcellation) of LeftInferiorParietal
M_ST24CV	-27.2158	1.77e-14	1.51e-12	1.30e-10	0.745	0.0526	9.14	1.00	2	Volume (Cortical Parcellation) of LeftEntorhinal
RD_ST52TA	1.5455	2.62e+00	4.69e+00	8.40e+00	0.742	0.0130	5.10	0.45	1	Cortical Thickness Average of LeftPrecuneus
ICV	-0.0337	9.56e-01	9.67e-01	9.78e-01	0.743	0.0144	5.78	1.00	1	NA
M_ST24SA	2.2443	3.49e+00	9.43e+00	2.55e+01	0.742	0.0135	4.34	0.25	2	Surface Area of LeftEntorhinal
M_ST12SV	-25.8479	4.60e-14	5.95e-12	7.70e-10	0.739	0.0425	7.86	1.00	2	Volume (WM Parcellation) of LeftAmygdala
M_ST39CV	14.8041	1.23e+04	2.69e+06	5.86e+08	0.740	0.0188	5.12	0.95	3	Volume (Cortical Parcellation) of LeftMedialOrbitofrontal
M_ST32CV	-13.7800	2.15e-08	1.04e-06	4.99e-05	0.738	0.0244	6.44	0.95	2	Volume (Cortical Parcellation) of LeftInferiorTemporal
M_ST46TS	116.2719	7.02e+29	3.14e+50	1.40e+71	0.741	0.0165	4.69	1.00	2	Cortical Thickness Standard Deviation of LeftParsOrbitalis
M_ST40CV	-7.7090	5.05e-05	4.49e-04	3.99e-03	0.733	0.0167	6.27	0.50	3	Volume (Cortical Parcellation) of LeftMiddleTemporal
M_ST39SA	3.0038	5.29e+00	2.02e+01	7.69e+01	0.744	0.0167	4.30	0.70	2	Surface Area of LeftMedialOrbitofrontal
M_ST32TA	-48.6612	7.79e-27	7.36e-22	6.95e-17	0.730	0.0323	6.97	0.70	2	Cortical Thickness Average of LeftInferiorTemporal
RD_ST32TA	0.4213	1.28e+00	1.52e+00	1.81e+00	0.731	0.0198	4.69	0.30	3	Cortical Thickness Average of LeftInferiorTemporal
Hippocampus	-9.9733	5.00e-06	4.66e-05	4.35e-04	0.721	0.0244	7.32	0.65	3	NA
M_ST40TA	-55.4204	1.29e-28	8.54e-25	5.64e-21	0.719	0.0564	9.38	0.45	4	Cortical Thickness Average of LeftMiddleTemporal
M_ST30SV	2.1112	4.98e+00	8.26e+00	1.37e+01	0.713	0.0314	6.62	0.25	4	Volume (WM Parcellation) of LeftInferiorLateralVentricle
M_ST32SA	-1.0856	2.38e-01	3.38e-01	4.78e-01	0.712	0.0243	5.65	0.25	4	Surface Area of LeftInferiorTemporal
M_ST45TA	22.3568	7.28e+05	5.12e+09	3.60e+13	0.716	0.0251	4.87	0.25	4	Cortical Thickness Average of LeftParsOpercularis

3.1 Decorrelating the sets

TADPOLE_DX_TRAIND <- GDSTMDecorrelation(TADPOLE_DX_TRAIN,Outcome="DX",
                                        thr=0.6,
                                        type="RLM",
                                        method="spearman",
                                        verbose = TRUE)
#> 
#>  Included: 197 , Uni p: 0.01215481 To Outcome: 99 , Base: 5 , In Included: 5 , Base Cor: 16 
#> 1 , Top: 37 < 0.6 >( 2 )[ 1 : 0 : 0.594 ]( 36 , 80 , 0 ),<>Tot Used: 116 , Added: 80 , Zero Std: 0 , Max Cor: 0.9913617 
#> 2 , Top: 22 < 0.6 >[ FALSE ]( 1 )[ 1 : 0 : 0.594 ]( 19 , 38 , 36 ),<>Tot Used: 143 , Added: 38 , Zero Std: 0 , Max Cor: 0.9278401 
#> 3 , Top: 19 < 0.6 >( 1 )[ 1 : 0 : 0.594 ]( 18 , 19 , 53 ),<>Tot Used: 152 , Added: 19 , Zero Std: 0 , Max Cor: 0.9040669 
#> 4 , Top: 9 < 0.6 >[ TRUE ]( 2 )[ 1 : 0 : 0 ]( 9 , 11 , 68 ),<>Tot Used: 155 , Added: 11 , Zero Std: 0 , Max Cor: 0.8911764 
#> 5 , Top: 6 < 0.6 >[ TRUE ]( 1 )[ 1 : 0 : 0 ]( 6 , 6 , 71 ),<>Tot Used: 155 , Added: 6 , Zero Std: 0 , Max Cor: 0.6234069 
#> 6 , Top: 3 < 0.6 >[ FALSE ]( 1 )[ 1 : 0 : 0 ]( 3 , 3 , 77 ),<>Tot Used: 158 , Added: 3 , Zero Std: 0 , Max Cor: 0.878083 
#> 7 , Top: 1 < 0.6 >[ FALSE ]( 1 )[ 1 : 0 : 0.6 ]( 1 , 1 , 80 ),<>Tot Used: 159 , Added: 1 , Zero Std: 0 , Max Cor: 0.5995342 
#> [ 8 ], 0.5995342 . Cor to Base: 88 , ABase: 69
TADPOLE_DX_TESTD <-  predictDecorrelate(TADPOLE_DX_TRAIND,TADPOLE_DX_TEST)

3.1.1 Decorrelated ML

bDXmlD <- BSWiMS.model(DX~.,TADPOLE_DX_TRAIND,NumberofRepeats = 20)

[++-++—++-+++-++-+–++-++-++—+++-++–+++-++-+++-+++-++-++-++-++-++-]….

pander::pander(bDXmlD$bagging$Jaccard.SM)

0.288

fs <- bDXmlD$bagging$frequencyTable
barplot(fs[order(-fs)],las=2,main="Selected Features",cex.names = 0.5)

sm <- summary(bDXmlD)
pander::pander(sm$coefficients)

	Estimate	lower	OR	upper	u.Accuracy	r.Accuracy	full.Accuracy	u.AUC	r.AUC	full.AUC	IDI	NRI	z.IDI	z.NRI	Frequency
Ba_M_ST24TA	-102.962	9.41e-52	1.92e-45	3.93e-39	0.722	0.635	0.755	0.723	0.638	0.760	0.1898	0.854	14.42	13.82	1.00
Ba_M_ST31TA	-200.910	1.04e-100	5.57e-88	2.97e-75	0.654	0.688	0.727	0.645	0.684	0.728	0.0979	0.551	9.75	8.41	0.95
De_M_ST26CV	-13.321	1.23e-07	1.64e-06	2.18e-05	0.635	0.694	0.734	0.634	0.697	0.739	0.0647	0.473	7.56	7.16	0.30
De_M_ST60CV	-80.738	1.67e-43	8.63e-36	4.47e-28	0.638	0.691	0.727	0.638	0.688	0.727	0.0581	0.515	7.48	7.81	0.90
De_M_ST48TA	291.460	1.56e+90	3.80e+126	9.23e+162	0.626	0.734	0.751	0.627	0.734	0.756	0.0428	0.478	6.39	7.21	1.00
De_M_ST44TA	-16.752	9.55e-10	5.31e-08	2.95e-06	0.625	0.700	0.724	0.626	0.699	0.727	0.0451	0.340	6.13	5.06	0.25
De_M_ST30SV	17.547	9.58e+04	4.18e+07	1.82e+10	0.637	0.719	0.727	0.628	0.721	0.728	0.0329	0.348	5.56	5.18	0.95
Ba_M_ST32SA	-1.318	1.65e-01	2.68e-01	4.33e-01	0.607	0.718	0.731	0.596	0.713	0.737	0.0328	0.389	5.30	5.80	0.20
De_M_ST31SA	-2.203	4.89e-02	1.10e-01	2.50e-01	0.548	0.713	0.722	0.556	0.712	0.724	0.0292	0.237	5.20	3.49	0.30
De_M_ST11SV	-42.720	1.82e-26	2.80e-19	4.29e-12	0.616	0.707	0.730	0.609	0.711	0.733	0.0268	0.308	5.01	4.56	0.20
M_ST129TS	117.340	2.46e+30	9.13e+50	3.39e+71	0.587	0.712	0.731	0.568	0.709	0.732	0.0265	0.323	4.76	4.80	0.45
De_M_ST43TA	13.297	2.50e+03	5.96e+05	1.42e+08	0.566	0.744	0.753	0.559	0.745	0.757	0.0256	0.305	4.69	4.52	0.10
RD_ST52TA	1.423	2.30e+00	4.15e+00	7.49e+00	0.609	0.738	0.749	0.558	0.745	0.752	0.0271	0.232	4.59	3.50	0.25
De_M_ST49CV	-27.259	1.23e-17	1.45e-12	1.70e-07	0.574	0.737	0.753	0.565	0.747	0.757	0.0236	0.150	4.50	2.20	0.95
M_ST43TS	146.322	2.09e+33	3.52e+63	5.94e+93	0.529	0.737	0.745	0.514	0.741	0.749	0.0198	0.201	4.00	2.96	0.50
Ba_M_ST13SA	-0.955	2.40e-01	3.85e-01	6.17e-01	0.581	0.713	0.723	0.595	0.710	0.715	0.0170	0.340	3.88	5.07	0.10

pander::pander(bDXmlD$univariate[bDXmlD$selectedfeatures,])

	Name	RName	ZUni
Ba_M_ST24TA	Ba_M_ST24TA	Ba_M_ST24TA	10.41
De_M_ST43TA	De_M_ST43TA	De_M_ST43TA	3.51
De_M_ST48TA	De_M_ST48TA	De_M_ST48TA	5.36
De_M_ST49CV	De_M_ST49CV	De_M_ST49CV	4.90
RD_ST52TA	RD_ST52TA	RD_ST52TA	3.90
Ba_M_ST31TA	Ba_M_ST31TA	Ba_M_ST31TA	7.28
De_M_ST60CV	De_M_ST60CV	De_M_ST60CV	5.86
De_M_ST30SV	De_M_ST30SV	De_M_ST30SV	7.03
M_ST43TS	M_ST43TS	M_ST43TS	2.26
M_ST129TS	M_ST129TS	M_ST129TS	3.24
De_M_ST26CV	De_M_ST26CV	De_M_ST26CV	6.34
De_M_ST31SA	De_M_ST31SA	De_M_ST31SA	4.01
Ba_M_ST13SA	Ba_M_ST13SA	Ba_M_ST13SA	3.62
De_M_ST62TA	De_M_ST62TA	De_M_ST62TA	2.71
De_M_ST44TA	De_M_ST44TA	De_M_ST44TA	6.73
Ba_M_ST32SA	Ba_M_ST32SA	Ba_M_ST32SA	5.06
De_M_ST11SV	De_M_ST11SV	De_M_ST11SV	5.54
De_M_ST32TA	De_M_ST32TA	De_M_ST32TA	4.08

prBin <- predictionStats_binary(cbind(TADPOLE_DX_TESTD$DX,predict(bDXmlD,TADPOLE_DX_TESTD)),"NL vs Dementia")

NL vs Dementia

pander::pander(prBin$aucs)

est	lower	upper
0.777	0.727	0.827

pander::pander(prBin$accc)

est	lower	upper
0.702	0.654	0.748

pander::pander(prBin$berror)

50%	2.5%	97.5%
0.322	0.272	0.378

pander::pander(prBin$sensitivity)

est	lower	upper
0.622	0.519	0.718

par(op)

3.1.2 The formula network

cmax <- apply(bDXmlD$bagging$formulaNetwork,2,max)
cnames <- names(cmax[cmax>=0.25])
cmax <- cmax[cnames]

adma <- bDXmlD$bagging$formulaNetwork[cnames,cnames]

adma[adma<0.15] <- 0;
gr <- graph_from_adjacency_matrix(adma,mode = "undirected",diag = FALSE,weighted=TRUE)
gr$layout <- layout_with_fr

fc <- cluster_optimal(gr)
plot(fc, gr,
     edge.width=10*E(gr)$weight,
     vertex.size=10*cmax,
     vertex.label.cex=0.75,
     vertex.label.dist=0,
     main="NL vs Dementia Diagnosis")

par(op)

3.1.3 Decorrelated MCI vs AD table

clusterFeatures <- fc$names
sm$coefficients$DeltaAUC <- (sm$coefficients$full.AUC-sm$coefficients$r.AUC)

tableMCI_DeD <- sm$coefficients[clusterFeatures,
                                   c("Estimate",
                                     "lower",
                                     "OR",
                                     "upper",
                                     "full.AUC",
                                     "DeltaAUC",
                                     "z.IDI",
                                     "Frequency")]

nugget <- fc$membership
names(nugget) <- clusterFeatures

tableMCI_DeD$Cluster <- nugget[rownames(tableMCI_DeD)]

rnames <- clusterFeatures[str_detect(clusterFeatures,"ST")]
frnames <- rnames
rnames <- str_replace_all(rnames,"M_","")
rnames <- str_replace_all(rnames,"RD_","")
rnames <- str_replace_all(rnames,"Ba_","")
rnames <- str_replace_all(rnames,"De_","")
description <- character()

for (ddet in c(1:length(rnames)))
{
  description <- c(description,TADPOLE_D1_D2_Dict$TEXT[str_detect(TADPOLE_D1_D2_Dict$FLDNAME,rnames[ddet])][1])
}
names(description) <- frnames

tableMCI_DeD$Description <- description[rownames(tableMCI_DeD)]



## Getting the decorrelation formula
dc <- getDerivedCoefficients(TADPOLE_DX_TRAIND)
decornames <- rownames(sm$coefficients)

deNames_in_dc <- decornames[decornames %in% names(dc)]
theDeFormulas <- dc[deNames_in_dc]
deFromula <- character(length(theDeFormulas))
names(deFromula) <- names(theDeFormulas)
for (dx in names(deFromula))
{
  coef <- theDeFormulas[[dx]]
  cname <- names(theDeFormulas[[dx]])
  names(cname) <- cname
  for (cf in names(coef))
  {
    if (cf != dx)
    {
      if (coef[cf]>0)
      {
        deFromula[dx] <- paste(deFromula[dx],
                               sprintf("+ %5.3f*%s",coef[cf],cname[cf]))
      }
      else
      {
        deFromula[dx] <- paste(deFromula[dx],
                               sprintf("%5.3f*%s",coef[cf],cname[cf]))
      }
    }
  }
}

tableMCI_DeD$DecorFormula <- deFromula[rownames(tableMCI_DeD)]



pander::pander(tableMCI_DeD)

	Estimate	lower	OR	upper	full.AUC	DeltaAUC	z.IDI	Frequency	Cluster	Description	DecorFormula
Ba_M_ST24TA	-102.96	9.41e-52	1.92e-45	3.93e-39	0.760	0.12195	14.42	1.00	1	Cortical Thickness Average of LeftEntorhinal	NA
De_M_ST48TA	291.46	1.56e+90	3.80e+126	9.23e+162	0.756	0.02211	6.39	1.00	1	Cortical Thickness Average of LeftPericalcarine	-0.425M_ST31TA + 1.000M_ST48TA -1.029*M_ST48TS
De_M_ST49CV	-27.26	1.23e-17	1.45e-12	1.70e-07	0.757	0.01015	4.50	0.95	1	Volume (Cortical Parcellation) of LeftPostcentral	-0.587M_ST39CV + 1.000M_ST49CV
RD_ST52TA	1.42	2.30e+00	4.15e+00	7.49e+00	0.752	0.00673	4.59	0.25	1	Cortical Thickness Average of LeftPrecuneus	NA
Ba_M_ST31TA	-200.91	1.04e-100	5.57e-88	2.97e-75	0.728	0.04417	9.75	0.95	2	Cortical Thickness Average of LeftInferiorParietal	NA
De_M_ST60CV	-80.74	1.67e-43	8.63e-36	4.47e-28	0.727	0.03921	7.48	0.90	2	Volume (Cortical Parcellation) of LeftTemporalPole	+ 0.758M_ST24TA -1.322M_ST60TA + 0.113M_ST39SA -0.335M_ST60SA -0.563M_ST39CV + 1.000M_ST60CV
De_M_ST30SV	17.55	9.58e+04	4.18e+07	1.82e+10	0.728	0.00745	5.56	0.95	2	Volume (WM Parcellation) of LeftInferiorLateralVentricle	-0.335Ventricles + 1.000M_ST30SV
M_ST43TS	146.32	2.09e+33	3.52e+63	5.94e+93	0.749	0.00819	4.00	0.50	1	Cortical Thickness Standard Deviation of LeftParacentral	NA
M_ST129TS	117.34	2.46e+30	9.13e+50	3.39e+71	0.732	0.02250	4.76	0.45	2	Cortical Thickness Standard Deviation of LeftInsula	NA
De_M_ST26CV	-13.32	1.23e-07	1.64e-06	2.18e-05	0.739	0.04238	7.56	0.30	1	Volume (Cortical Parcellation) of LeftFusiform	-0.219M_ST26SA + 1.000M_ST26CV -0.667*M_ST39CV
De_M_ST31SA	-2.20	4.89e-02	1.10e-01	2.50e-01	0.724	0.01241	5.20	0.30	2	Surface Area of LeftInferiorParietal	NA
De_M_ST44TA	-16.75	9.55e-10	5.31e-08	2.95e-06	0.727	0.02849	6.13	0.25	1	Cortical Thickness Average of LeftParahippocampal	NA

4 Diagnosis NL vs AD

4.0.1 the set

TrainFraction <- 0.60;

table(validBaselineTadpole$DX)

Dementia MCI NL 244 711 452

TADPOLECrossMRI <- subset(validBaselineTadpole,DX == "Dementia" | DX == "NL")
table(TADPOLECrossMRI$DX)

Dementia MCI NL 244 0 452

TADPOLECrossMRI$DX <- 1*(as.character(TADPOLECrossMRI$DX) == "Dementia")
table(TADPOLECrossMRI$DX)

0 1 452 244

TADPOLECrossMRI$ADAS13 <- NULL
TADPOLECrossMRI$ADAS11 <- NULL
TADPOLECrossMRI$MMSE <- NULL
TADPOLECrossMRI$RAVLT_immediate <- NULL
TADPOLECrossMRI$RAVLT_learning <- NULL
TADPOLECrossMRI$RAVLT_perc_forgetting <- NULL
TADPOLECrossMRI$RAVLT_forgetting <- NULL
TADPOLECrossMRI$FAQ <- NULL

TADPOLE_Cases <- subset(TADPOLECrossMRI,DX==1)
TADPOLE_Controls <- subset(TADPOLECrossMRI,DX==0)
trainCasesSet <- sample(nrow(TADPOLE_Cases),nrow(TADPOLE_Cases)*TrainFraction)
trainControlSet <- sample(nrow(TADPOLE_Controls),nrow(TADPOLE_Controls)*TrainFraction)

TADPOLE_DX_NLDE_TRAIN <- rbind(TADPOLE_Cases[trainCasesSet,],TADPOLE_Controls[trainControlSet,])
TADPOLE_DX_NLDE_TEST <- TADPOLECrossMRI[!(rownames(TADPOLECrossMRI) %in% rownames(TADPOLE_DX_NLDE_TRAIN)),]


pander::pander(table(TADPOLE_DX_NLDE_TRAIN$DX))

0	1
271	146

pander::pander(table(TADPOLE_DX_NLDE_TEST$DX))

0	1
181	98

par(op)

4.0.2 Learning

bDXmlNLDE <- BSWiMS.model(DX~.,TADPOLE_DX_NLDE_TRAIN,NumberofRepeats = 20)

[++–+—+-+-+++-+++-+++-+++-+++-++–+–+—+++-++-++–+—++-++–+–+++-+–]….

pander::pander(bDXmlNLDE$bagging$Jaccard.SM)

0.273

fs <- bDXmlNLDE$bagging$frequencyTable
barplot(fs[order(-fs)],las=2,main="Selected Features",cex.names = 0.5)

sm <- summary(bDXmlNLDE)
pander::pander(sm$coefficients)

	Estimate	lower	OR	upper	u.Accuracy	r.Accuracy	full.Accuracy	u.AUC	r.AUC	full.AUC	IDI	NRI	z.IDI	z.NRI	Frequency
M_ST30SV	21.54	1.69e+07	2.26e+09	3.01e+11	0.759	0.841	0.871	0.757	0.836	0.868	0.0782	0.760	7.11	9.76	0.40
Hippocampus	-81.07	8.64e-45	6.17e-36	4.40e-27	0.821	0.871	0.900	0.816	0.868	0.899	0.0750	0.791	6.83	10.23	1.00
RD_ST32TA	13.02	1.27e+04	4.49e+05	1.59e+07	0.684	0.867	0.900	0.666	0.863	0.899	0.0745	0.766	6.82	9.85	1.00
M_ST24TA	-159.96	2.93e-90	3.39e-70	3.92e-50	0.826	0.873	0.899	0.818	0.872	0.899	0.0663	0.700	6.41	8.86	1.00
M_ST44CV	-52.54	7.80e-31	1.53e-23	3.00e-16	0.759	0.842	0.871	0.749	0.838	0.868	0.0619	0.776	6.04	9.98	0.40
M_ST12SV	-46.93	8.03e-27	4.17e-21	2.17e-15	0.826	0.865	0.889	0.823	0.860	0.886	0.0479	0.568	5.92	7.03	0.65
M_ST24CV	-38.45	5.48e-23	2.00e-17	7.31e-12	0.820	0.865	0.890	0.814	0.866	0.887	0.0549	0.560	5.82	6.97	0.65
M_ST51CV	13.08	4.76e+03	4.78e+05	4.80e+07	0.579	0.867	0.901	0.579	0.867	0.898	0.0448	0.509	5.49	6.23	0.30
M_ST40TA	-66.33	9.70e-40	1.57e-29	2.53e-19	0.762	0.849	0.870	0.754	0.847	0.867	0.0503	0.562	5.42	6.92	0.20
M_ST32CV	-19.49	2.61e-12	3.43e-09	4.50e-06	0.759	0.850	0.870	0.753	0.846	0.868	0.0468	0.684	5.24	8.58	0.30
M_ST13TA	-44.02	4.07e-27	7.66e-20	1.44e-12	0.722	0.864	0.876	0.715	0.862	0.873	0.0465	0.509	5.00	6.23	0.10
M_ST11SV	140.83	4.57e+36	1.45e+61	4.61e+85	0.582	0.854	0.872	0.582	0.853	0.870	0.0354	0.539	4.72	6.62	0.25
Ventricles	-2.67	2.32e-02	6.91e-02	2.06e-01	0.670	0.849	0.868	0.671	0.844	0.865	0.0348	0.651	4.68	8.23	0.15
M_ST31TA	-112.71	1.85e-70	1.13e-49	6.85e-29	0.705	0.878	0.893	0.699	0.874	0.890	0.0346	0.457	4.48	5.62	0.40
M_ST62TA	79.35	5.72e+19	2.88e+34	1.45e+49	0.594	0.859	0.871	0.588	0.855	0.869	0.0366	0.588	4.43	7.29	0.30
M_ST40CV	-29.80	2.58e-19	1.15e-13	5.12e-08	0.741	0.873	0.891	0.742	0.869	0.889	0.0353	0.536	4.37	6.54	0.60
RD_ST40TA	3.18	5.94e+00	2.40e+01	9.72e+01	0.632	0.854	0.873	0.608	0.851	0.869	0.0345	0.411	4.27	4.99	0.40
M_ST45TA	27.30	3.23e+06	7.17e+11	1.59e+17	0.620	0.852	0.872	0.608	0.852	0.870	0.0337	0.571	4.25	7.03	0.10
M_ST38TA	48.21	1.74e+11	8.63e+20	4.29e+30	0.582	0.848	0.869	0.583	0.847	0.864	0.0329	0.595	4.20	7.35	0.10
M_ST43TS	637.16	2.32e+148	5.21e+276	Inf	0.543	0.892	0.900	0.537	0.890	0.899	0.0299	0.570	4.12	7.09	0.95
M_ST39CV	3.27	5.58e+00	2.62e+01	1.23e+02	0.580	0.867	0.896	0.577	0.867	0.899	0.0254	0.263	4.11	3.11	0.10
M_ST24TS	77.78	3.74e+16	5.99e+33	9.61e+50	0.620	0.888	0.896	0.621	0.884	0.895	0.0200	0.254	3.75	3.00	0.45
M_ST59SA	-10.02	2.63e-07	4.47e-05	7.59e-03	0.560	0.897	0.900	0.567	0.895	0.900	0.0243	0.342	3.72	4.09	0.65

pander::pander(bDXmlNLDE$univariate[bDXmlNLDE$selectedfeatures,])

	Name	RName	ZUni
RD_ST32TA	RD_ST32TA	RD_ST32TA	8.21
Hippocampus	Hippocampus	Hippocampus	17.66
M_ST24TA	M_ST24TA	M_ST24TA	17.25
M_ST43TS	M_ST43TS	M_ST43TS	1.92
M_ST59SA	M_ST59SA	M_ST59SA	4.00
M_ST12SV	M_ST12SV	M_ST12SV	16.99
M_ST24CV	M_ST24CV	M_ST24CV	16.30
RD_ST55TA	RD_ST55TA	RD_ST55TA	2.87
M_ST40CV	M_ST40CV	M_ST40CV	13.21
M_ST60CV	M_ST60CV	M_ST60CV	6.61
M_ST24TS	M_ST24TS	M_ST24TS	5.70
M_ST60SA	M_ST60SA	M_ST60SA	2.80
M_ST39CV	M_ST39CV	M_ST39CV	4.03
M_ST51CV	M_ST51CV	M_ST51CV	4.19
M_ST44CV	M_ST44CV	M_ST44CV	11.50
M_ST32CV	M_ST32CV	M_ST32CV	13.38
M_ST31TA	M_ST31TA	M_ST31TA	10.24
RD_ST40TA	RD_ST40TA	RD_ST40TA	6.46
M_ST30SV	M_ST30SV	M_ST30SV	12.46
M_ST43TA	M_ST43TA	M_ST43TA	3.44
M_ST60TS	M_ST60TS	M_ST60TS	4.77
M_ST62TA	M_ST62TA	M_ST62TA	3.95
M_ST47TS	M_ST47TS	M_ST47TS	2.35
M_ST40TA	M_ST40TA	M_ST40TA	12.66
Ventricles	Ventricles	Ventricles	7.49
M_ST38TA	M_ST38TA	M_ST38TA	5.42
M_ST45TA	M_ST45TA	M_ST45TA	4.90
M_ST11SV	M_ST11SV	M_ST11SV	5.06
M_ST13TA	M_ST13TA	M_ST13TA	10.40

prBin <- predictionStats_binary(cbind(TADPOLE_DX_NLDE_TEST$DX,predict(bDXmlNLDE,TADPOLE_DX_NLDE_TEST)),"NL vs Dementia")

NL vs Dementia

pander::pander(prBin$aucs)

est	lower	upper
0.945	0.92	0.971

pander::pander(prBin$accc)

est	lower	upper
0.871	0.826	0.908

pander::pander(prBin$berror)

50%	2.5%	97.5%
0.136	0.0941	0.182

pander::pander(prBin$sensitivity)

est	lower	upper
0.837	0.748	0.904

par(op)

4.0.3 The formula network

cmax <- apply(bDXmlNLDE$bagging$formulaNetwork,2,max)
cnames <- names(cmax[cmax>=0.25])
cmax <- cmax[cnames]

adma <- bDXmlNLDE$bagging$formulaNetwork[cnames,cnames]

adma[adma<0.15] <- 0;
gr <- graph_from_adjacency_matrix(adma,mode = "undirected",diag = FALSE,weighted=TRUE)
gr$layout <- layout_with_fr

fc <- cluster_optimal(gr)
plot(fc, gr,
     edge.width=10*E(gr)$weight,
     vertex.size=10*cmax,
     vertex.label.cex=0.75,
     vertex.label.dist=0,
     main="NL vs Dementia Diagnosis")

par(op)

4.0.4 NL vs Dementia table

clusterFeatures <- fc$names
sm$coefficients$DeltaAUC <- (sm$coefficients$full.AUC-sm$coefficients$r.AUC)

tableNL_DE <- sm$coefficients[clusterFeatures,
                                   c("Estimate",
                                     "lower",
                                     "OR",
                                     "upper",
                                     "full.AUC",
                                     "DeltaAUC",
                                     "z.IDI",
                                     "Frequency")]

nugget <- fc$membership
names(nugget) <- clusterFeatures

tableNL_DE$Cluster <- nugget[rownames(tableNL_DE)]

rnames <- clusterFeatures[str_detect(clusterFeatures,"ST")]
frnames <- rnames
rnames <- str_replace_all(rnames,"M_","")
rnames <- str_replace_all(rnames,"RD_","")
description <- character()

for (ddet in c(1:length(rnames)))
{
  description <- c(description,TADPOLE_D1_D2_Dict$TEXT[str_detect(TADPOLE_D1_D2_Dict$FLDNAME,rnames[ddet])][1])
}
names(description) <- frnames

tableNL_DE$Description <- description[rownames(tableNL_DE)]
pander::pander(tableNL_DE)

	Estimate	lower	OR	upper	full.AUC	DeltaAUC	z.IDI	Frequency	Cluster	Description
RD_ST32TA	13.02	1.27e+04	4.49e+05	1.59e+07	0.899	0.03591	6.82	1.00	1	Cortical Thickness Average of LeftInferiorTemporal
Hippocampus	-81.07	8.64e-45	6.17e-36	4.40e-27	0.899	0.03152	6.83	1.00	1	NA
M_ST24TA	-159.96	2.93e-90	3.39e-70	3.92e-50	0.899	0.02604	6.41	1.00	1	Cortical Thickness Average of LeftEntorhinal
M_ST43TS	637.16	2.32e+148	5.21e+276	Inf	0.899	0.00982	4.12	0.95	1	Cortical Thickness Standard Deviation of LeftParacentral
M_ST59SA	-10.02	2.63e-07	4.47e-05	7.59e-03	0.900	0.00494	3.72	0.65	1	Surface Area of LeftSupramarginal
M_ST12SV	-46.93	8.03e-27	4.17e-21	2.17e-15	0.886	0.02540	5.92	0.65	2	Volume (WM Parcellation) of LeftAmygdala
M_ST24CV	-38.45	5.48e-23	2.00e-17	7.31e-12	0.887	0.02128	5.82	0.65	2	Volume (Cortical Parcellation) of LeftEntorhinal
M_ST40CV	-29.80	2.58e-19	1.15e-13	5.12e-08	0.889	0.01966	4.37	0.60	2	Volume (Cortical Parcellation) of LeftMiddleTemporal
M_ST24TS	77.78	3.74e+16	5.99e+33	9.61e+50	0.895	0.01043	3.75	0.45	2	Cortical Thickness Standard Deviation of LeftEntorhinal
M_ST51CV	13.08	4.76e+03	4.78e+05	4.80e+07	0.898	0.03077	5.49	0.30	2	Volume (Cortical Parcellation) of LeftPrecentral
M_ST44CV	-52.54	7.80e-31	1.53e-23	3.00e-16	0.868	0.02959	6.04	0.40	3	Volume (Cortical Parcellation) of LeftParahippocampal
M_ST32CV	-19.49	2.61e-12	3.43e-09	4.50e-06	0.868	0.02203	5.24	0.30	3	Volume (Cortical Parcellation) of LeftInferiorTemporal
M_ST31TA	-112.71	1.85e-70	1.13e-49	6.85e-29	0.890	0.01511	4.48	0.40	2	Cortical Thickness Average of LeftInferiorParietal
RD_ST40TA	3.18	5.94e+00	2.40e+01	9.72e+01	0.869	0.01772	4.27	0.40	3	Cortical Thickness Average of LeftMiddleTemporal
M_ST30SV	21.54	1.69e+07	2.26e+09	3.01e+11	0.868	0.03185	7.11	0.40	3	Volume (WM Parcellation) of LeftInferiorLateralVentricle
M_ST62TA	79.35	5.72e+19	2.88e+34	1.45e+49	0.869	0.01307	4.43	0.30	3	Cortical Thickness Average of LeftTransverseTemporal
M_ST11SV	140.83	4.57e+36	1.45e+61	4.61e+85	0.870	0.01658	4.72	0.25	3	Volume (WM Parcellation) of LeftAccumbensArea

4.1 Decorrelated Set

TADPOLE_DX_NLDE_TRAIND <- GDSTMDecorrelation(TADPOLE_DX_NLDE_TRAIN,Outcome="DX",
                                        thr=0.6,
                                        type="RLM",
                                        method="spearman",
                                        verbose = TRUE)
#> 
#>  Included: 194 , Uni p: 0.0110018 To Outcome: 109 , Base: 5 , In Included: 5 , Base Cor: 18 
#> 1 , Top: 40 < 0.6 >( 2 )[ 1 : 0 : 0.594 ]( 40 , 83 , 0 ),<>Tot Used: 123 , Added: 83 , Zero Std: 0 , Max Cor: 0.9669697 
#> 2 , Top: 17 < 0.6 >( 1 )[ 1 : 0 : 0.594 ]( 16 , 28 , 40 ),<>Tot Used: 140 , Added: 28 , Zero Std: 0 , Max Cor: 0.9627667 
#> 3 , Top: 15 < 0.6 >[ FALSE ]( 1 )[ 1 : 0 : 0.594 ]( 14 , 22 , 52 ),<>Tot Used: 152 , Added: 22 , Zero Std: 0 , Max Cor: 0.9130056 
#> 4 , Top: 14 < 0.6 >( 1 )[ 1 : 0 : 0 ]( 13 , 14 , 62 ),<>Tot Used: 162 , Added: 14 , Zero Std: 0 , Max Cor: 0.9517803 
#> 5 , Top: 5 < 0.6 >( 1 )[ 1 : 0 : 0 ]( 5 , 5 , 72 ),<>Tot Used: 164 , Added: 5 , Zero Std: 0 , Max Cor: 0.8156711 
#> 6 , Top: 1 < 0.6 >( 1 )[ 1 : 0 : 0.6 ]( 1 , 1 , 76 ),<>Tot Used: 164 , Added: 1 , Zero Std: 0 , Max Cor: 0.5978659 
#> [ 7 ], 0.5978659 . Cor to Base: 91 , ABase: 65
TADPOLE_DX_NLDE_TESTD <-  predictDecorrelate(TADPOLE_DX_NLDE_TRAIND,TADPOLE_DX_NLDE_TEST)

4.1.1 Learning

bDXmlNLDED <- BSWiMS.model(DX~.,TADPOLE_DX_NLDE_TRAIND,NumberofRepeats = 20)

[+++-++-++-+++-++–++-++-++-++-++-++-+++-++-+++-++-++-+++-++-++-++-]….

pander::pander(bDXmlNLDED$bagging$Jaccard.SM)

0.294

fs <- bDXmlNLDED$bagging$frequencyTable
barplot(fs[order(-fs)],las=2,main="Selected Features",cex.names = 0.5)

sm <- summary(bDXmlNLDED)
pander::pander(sm$coefficients)

	Estimate	lower	OR	upper	u.Accuracy	r.Accuracy	full.Accuracy	u.AUC	r.AUC	full.AUC	IDI	NRI	z.IDI	z.NRI	Frequency
Ba_Hippocampus	-111.406	6.16e-56	4.14e-49	2.79e-42	0.822	0.714	0.905	0.816	0.701	0.901	0.4362	1.457	21.23	25.11	1.00
M_ST44SA	-42.442	9.26e-22	3.69e-19	1.48e-16	0.569	0.789	0.877	0.572	0.787	0.874	0.2176	1.169	12.72	16.93	0.25
De_M_ST44CV	109.399	1.65e+40	3.25e+47	6.40e+54	0.555	0.815	0.881	0.549	0.812	0.877	0.1634	0.964	10.55	13.45	0.30
M_ST44TA	-292.362	3.07e-156	1.07e-127	3.72e-99	0.703	0.821	0.895	0.697	0.816	0.891	0.1492	0.806	9.74	11.33	0.90
Ba_M_ST31TA	-316.526	1.01e-165	3.42e-138	1.16e-110	0.706	0.850	0.898	0.699	0.847	0.895	0.1167	0.910	8.36	12.26	1.00
De_M_ST32CV	-27.349	1.66e-15	1.33e-12	1.06e-09	0.575	0.864	0.898	0.580	0.858	0.894	0.0745	0.646	7.07	7.96	0.15
De_M_ST51TA	752.294	5.11e+248	Inf	Inf	0.626	0.881	0.905	0.609	0.878	0.901	0.0810	0.691	6.82	8.80	1.00
De_M_ST24TA	-123.451	1.73e-70	2.43e-54	3.43e-38	0.626	0.880	0.905	0.619	0.875	0.901	0.0594	0.641	6.29	8.01	1.00
De_M_ST58CV	195.220	3.96e+59	6.07e+84	9.30e+109	0.675	0.869	0.898	0.674	0.863	0.895	0.0613	0.732	6.05	9.29	1.00
De_M_ST32TA	63.144	4.41e+17	2.65e+27	1.59e+37	0.548	0.873	0.898	0.547	0.866	0.894	0.0481	0.625	5.47	7.69	0.15
De_Ventricles	138.846	6.31e+37	2.00e+60	6.31e+82	0.726	0.882	0.898	0.711	0.879	0.895	0.0405	0.472	5.14	5.79	1.00
De_M_ST24CV	-22.826	2.14e-14	1.22e-10	6.96e-07	0.612	0.858	0.878	0.610	0.858	0.874	0.0414	0.485	5.11	5.93	0.20
De_M_ST38TA	172.712	2.96e+46	1.02e+75	3.51e+103	0.572	0.871	0.889	0.560	0.868	0.885	0.0439	0.507	5.02	6.17	0.30
RD_ST32TA	6.508	5.65e+01	6.71e+02	7.96e+03	0.684	0.886	0.903	0.667	0.882	0.899	0.0402	0.590	5.02	7.26	1.00
De_M_ST13TA	-68.702	4.38e-42	1.46e-30	4.84e-19	0.575	0.868	0.889	0.570	0.862	0.886	0.0404	0.612	4.96	7.60	0.20
M_ST44TS	-75.292	1.40e-46	2.00e-33	2.85e-20	0.524	0.867	0.889	0.533	0.864	0.886	0.0345	0.672	4.80	8.41	0.20
Ba_M_ST32SA	-9.750	9.62e-07	5.83e-05	3.53e-03	0.678	0.880	0.898	0.669	0.874	0.895	0.0303	0.485	4.48	5.91	1.00
RD_ST15TA	4.170	9.76e+00	6.47e+01	4.29e+02	0.598	0.894	0.905	0.569	0.890	0.901	0.0276	0.433	4.21	5.34	0.60
RD_ST40TA	6.239	3.01e+01	5.13e+02	8.74e+03	0.631	0.886	0.898	0.607	0.882	0.894	0.0303	0.449	4.20	5.48	0.60
De_M_ST12SV	-11.916	2.71e-08	6.68e-06	1.65e-03	0.558	0.871	0.884	0.564	0.866	0.881	0.0242	0.357	4.19	4.27	0.10
De_M_ST43TA	215.897	3.54e+49	5.79e+93	9.49e+137	0.593	0.886	0.896	0.591	0.884	0.893	0.0305	0.496	3.97	6.05	0.90
RD_ST47TS	-0.485	4.78e-01	6.15e-01	7.92e-01	0.509	0.878	0.892	0.530	0.873	0.885	0.0156	0.328	3.73	3.92	0.10
Ba_M_ST59SA	-4.561	8.52e-04	1.05e-02	1.28e-01	0.561	0.900	0.902	0.569	0.892	0.899	0.0210	0.296	3.46	3.52	0.60

pander::pander(bDXmlNLDED$univariate[bDXmlNLDED$selectedfeatures,])

	Name	RName	ZUni
Ba_Hippocampus	Ba_Hippocampus	Ba_Hippocampus	17.66
De_M_ST51TA	De_M_ST51TA	De_M_ST51TA	6.49
De_M_ST24TA	De_M_ST24TA	De_M_ST24TA	5.33
RD_ST15TA	RD_ST15TA	RD_ST15TA	3.62
Ba_M_ST31TA	Ba_M_ST31TA	Ba_M_ST31TA	10.24
RD_ST32TA	RD_ST32TA	RD_ST32TA	8.21
De_Ventricles	De_Ventricles	De_Ventricles	9.48
De_M_ST58CV	De_M_ST58CV	De_M_ST58CV	8.26
Ba_M_ST32SA	Ba_M_ST32SA	Ba_M_ST32SA	8.14
RD_ST40TA	RD_ST40TA	RD_ST40TA	6.46
De_M_ST32CV	De_M_ST32CV	De_M_ST32CV	3.00
M_ST44TS	M_ST44TS	M_ST44TS	1.69
M_ST44TA	M_ST44TA	M_ST44TA	9.64
Ba_M_ST59SA	Ba_M_ST59SA	Ba_M_ST59SA	4.00
De_M_ST13TA	De_M_ST13TA	De_M_ST13TA	3.47
De_M_ST32TA	De_M_ST32TA	De_M_ST32TA	2.99
De_M_ST43TA	De_M_ST43TA	De_M_ST43TA	4.78
De_M_ST38TA	De_M_ST38TA	De_M_ST38TA	2.22
De_M_ST44CV	De_M_ST44CV	De_M_ST44CV	2.53
M_ST44SA	M_ST44SA	M_ST44SA	3.82
De_M_ST12SV	De_M_ST12SV	De_M_ST12SV	3.31
De_M_ST24CV	De_M_ST24CV	De_M_ST24CV	5.21
RD_ST47TS	RD_ST47TS	RD_ST47TS	1.88
De_M_ST40CV	De_M_ST40CV	De_M_ST40CV	3.80
De_M_ST45TA	De_M_ST45TA	De_M_ST45TA	2.89

prBin <- predictionStats_binary(cbind(TADPOLE_DX_NLDE_TESTD$DX,predict(bDXmlNLDED,TADPOLE_DX_NLDE_TESTD)),"NL vs Dementia")

NL vs Dementia

pander::pander(prBin$aucs)

est	lower	upper
0.951	0.925	0.977

pander::pander(prBin$accc)

est	lower	upper
0.882	0.838	0.917

pander::pander(prBin$berror)

50%	2.5%	97.5%
0.118	0.0803	0.16

pander::pander(prBin$sensitivity)

est	lower	upper
0.878	0.796	0.935

par(op)

4.1.2 The formula network

cmax <- apply(bDXmlNLDED$bagging$formulaNetwork,2,max)
cnames <- names(cmax[cmax>=0.25])
cmax <- cmax[cnames]

adma <- bDXmlNLDED$bagging$formulaNetwork[cnames,cnames]

adma[adma<0.15] <- 0;
gr <- graph_from_adjacency_matrix(adma,mode = "undirected",diag = FALSE,weighted=TRUE)
gr$layout <- layout_with_fr

fc <- cluster_optimal(gr)
plot(fc, gr,
     edge.width=10*E(gr)$weight,
     vertex.size=10*cmax,
     vertex.label.cex=0.75,
     vertex.label.dist=0,
     main="NL vs Dementia Diagnosis")

par(op)

4.1.3 Decorrelated NL vs Dementia table

clusterFeatures <- fc$names
sm$coefficients$DeltaAUC <- (sm$coefficients$full.AUC-sm$coefficients$r.AUC)

tableNL_DED <- sm$coefficients[clusterFeatures,
                                   c("Estimate",
                                     "lower",
                                     "OR",
                                     "upper",
                                     "full.AUC",
                                     "DeltaAUC",
                                     "z.IDI",
                                     "Frequency")]

nugget <- fc$membership
names(nugget) <- clusterFeatures

tableNL_DED$Cluster <- nugget[rownames(tableNL_DED)]

rnames <- clusterFeatures[str_detect(clusterFeatures,"ST")]
frnames <- rnames
rnames <- str_replace_all(rnames,"M_","")
rnames <- str_replace_all(rnames,"RD_","")
rnames <- str_replace_all(rnames,"Ba_","")
rnames <- str_replace_all(rnames,"De_","")
description <- character()

for (ddet in c(1:length(rnames)))
{
  description <- c(description,TADPOLE_D1_D2_Dict$TEXT[str_detect(TADPOLE_D1_D2_Dict$FLDNAME,rnames[ddet])][1])
}
names(description) <- frnames

tableNL_DED$Description <- description[rownames(tableNL_DED)]



## Getting the decorrelation formula
dc <- getDerivedCoefficients(TADPOLE_DX_NLDE_TRAIND)
decornames <- rownames(sm$coefficients)

deNames_in_dc <- decornames[decornames %in% names(dc)]
theDeFormulas <- dc[deNames_in_dc]
deFromula <- character(length(theDeFormulas))
names(deFromula) <- names(theDeFormulas)
for (dx in names(deFromula))
{
  coef <- theDeFormulas[[dx]]
  cname <- names(theDeFormulas[[dx]])
  names(cname) <- cname
  for (cf in names(coef))
  {
    if (cf != dx)
    {
      if (coef[cf]>0)
      {
        deFromula[dx] <- paste(deFromula[dx],
                               sprintf("+ %5.3f*%s",coef[cf],cname[cf]))
      }
      else
      {
        deFromula[dx] <- paste(deFromula[dx],
                               sprintf("%5.3f*%s",coef[cf],cname[cf]))
      }
    }
  }
}

tableNL_DED$DecorFormula <- deFromula[rownames(tableNL_DED)]



pander::pander(tableNL_DED)

	Estimate	lower	OR	upper	full.AUC	DeltaAUC	z.IDI	Frequency	Cluster	Description	DecorFormula
Ba_Hippocampus	-111.41	6.16e-56	4.14e-49	2.79e-42	0.901	0.20002	21.23	1.00	1	NA	NA
De_M_ST51TA	752.29	5.11e+248	Inf	Inf	0.901	0.02285	6.82	1.00	1	Cortical Thickness Average of LeftPrecentral	-0.507M_ST31TA + 1.000M_ST51TA + 0.032M_ST51SA -0.140M_ST51CV
De_M_ST24TA	-123.45	1.73e-70	2.43e-54	3.43e-38	0.901	0.02584	6.29	1.00	1	Cortical Thickness Average of LeftEntorhinal	-0.324Hippocampus + 1.000M_ST24TA
RD_ST15TA	4.17	9.76e+00	6.47e+01	4.29e+02	0.901	0.01102	4.21	0.60	1	Cortical Thickness Average of LeftCaudalMiddleFrontal	NA
Ba_M_ST31TA	-316.53	1.01e-165	3.42e-138	1.16e-110	0.895	0.04809	8.36	1.00	2	Cortical Thickness Average of LeftInferiorParietal	NA
RD_ST32TA	6.51	5.65e+01	6.71e+02	7.96e+03	0.899	0.01711	5.02	1.00	2	Cortical Thickness Average of LeftInferiorTemporal	NA
De_Ventricles	138.85	6.31e+37	2.00e+60	6.31e+82	0.895	0.01522	5.14	1.00	2	NA	+ 1.000Ventricles -1.285M_ST37SV
De_M_ST58CV	195.22	3.96e+59	6.07e+84	9.30e+109	0.895	0.03127	6.05	1.00	2	Volume (Cortical Parcellation) of LeftSuperiorTemporal	-0.125Hippocampus -2.854M_ST58TA -0.212M_ST58SA + 1.000M_ST58CV
Ba_M_ST32SA	-9.75	9.62e-07	5.83e-05	3.53e-03	0.895	0.02019	4.48	1.00	2	Surface Area of LeftInferiorTemporal	NA
RD_ST40TA	6.24	3.01e+01	5.13e+02	8.74e+03	0.894	0.01181	4.20	0.60	1	Cortical Thickness Average of LeftMiddleTemporal	NA
M_ST44TA	-292.36	3.07e-156	1.07e-127	3.72e-99	0.891	0.07512	9.74	0.90	2	Cortical Thickness Average of LeftParahippocampal	NA
Ba_M_ST59SA	-4.56	8.52e-04	1.05e-02	1.28e-01	0.899	0.00683	3.46	0.60	2	Surface Area of LeftSupramarginal	NA
De_M_ST43TA	215.90	3.54e+49	5.79e+93	9.49e+137	0.893	0.00880	3.97	0.90	2	Cortical Thickness Average of LeftParacentral	-0.708M_ST31TA + 1.000M_ST43TA
De_M_ST38TA	172.71	2.96e+46	1.02e+75	3.51e+103	0.885	0.01643	5.02	0.30	3	Cortical Thickness Average of LeftLingual	-0.550M_ST31TA + 1.000M_ST38TA
De_M_ST44CV	109.40	1.65e+40	3.25e+47	6.40e+54	0.877	0.06500	10.55	0.30	3	Volume (Cortical Parcellation) of LeftParahippocampal	-0.380Hippocampus + 1.000M_ST44CV
M_ST44SA	-42.44	9.26e-22	3.69e-19	1.48e-16	0.874	0.08626	12.72	0.25	3	Surface Area of LeftParahippocampal	NA

5 Prognosis MCI to AD Conversion

5.1 the set

TrainFraction <- 0.60;

MCIPrognosisIDs <- c(MCIconverters$PTID,MCI_No_converters$PTID)

TADPOLECrossMRI <- validBaselineTadpole[MCIPrognosisIDs,]
table(TADPOLECrossMRI$DX)

Dementia MCI NL 0 680 0

TADPOLECrossMRI$DX <- NULL
TADPOLECrossMRI$status <- 1*(rownames(TADPOLECrossMRI) %in% MCIconverters$PTID)
table(TADPOLECrossMRI$status)

0 1 436 244

TADPOLECrossMRI$TimeToEvent <- numeric(nrow(TADPOLECrossMRI))
TADPOLECrossMRI[MCIconverters$PTID,"TimeToEvent"] <- MCIconverters$TimeToEvent
TADPOLECrossMRI[MCI_No_converters$PTID,"TimeToEvent"] <- MCI_No_converters$TimeToEvent

TADPOLE_Cases <- subset(TADPOLECrossMRI,status==1)
TADPOLE_Controls <- subset(TADPOLECrossMRI,status==0)
trainCasesSet <- sample(nrow(TADPOLE_Cases),nrow(TADPOLE_Cases)*TrainFraction)
trainControlSet <- sample(nrow(TADPOLE_Controls),nrow(TADPOLE_Controls)*TrainFraction)

TADPOLE_Conv_TRAIN <- rbind(TADPOLE_Cases[trainCasesSet,],TADPOLE_Controls[trainControlSet,])
TADPOLE_Conv_TEST <- TADPOLECrossMRI[!(rownames(TADPOLECrossMRI) %in%
                                         rownames(TADPOLE_Conv_TRAIN)),]

pander::pander(table(TADPOLE_Conv_TRAIN$status))

0	1
261	146

pander::pander(table(TADPOLE_Conv_TEST$status))

0	1
175	98

par(op)

5.1.1 Learning

bConvml <- BSWiMS.model(Surv(TimeToEvent,status)~1,TADPOLE_Conv_TRAIN,NumberofRepeats = 20)

[++-+++-++-++-+–+–+–+++-++-++–++-+++-+–++-+–++-++-++-+–+–]…

pander::pander(bConvml$bagging$Jaccard.SM)

0.267

fs <- bConvml$bagging$frequencyTable
barplot(fs[order(-fs)],las=2,main="Selected Features",cex.names = 0.5)

sm <- summary(bConvml)
pander::pander(sm$coefficients)

	Estimate	lower	HR	upper	u.Accuracy	r.Accuracy	full.Accuracy	u.AUC	r.AUC	full.AUC	IDI	NRI	z.IDI	z.NRI	Frequency
FAQ	0.06129	1.05e+00	1.06e+00	1.08e+00	0.732	0.736	0.772	0.701	0.742	0.774	0.03870	0.500	6.51	6.15	1.00
M_ST12SV	-5.66104	6.27e-04	3.48e-03	1.93e-02	0.678	0.700	0.725	0.675	0.701	0.727	0.04227	0.524	6.42	6.23	0.15
ADAS13	0.06511	1.05e+00	1.07e+00	1.09e+00	0.702	0.749	0.772	0.698	0.755	0.774	0.04813	0.401	5.86	4.74	1.00
RD_ST24CV	0.63042	1.49e+00	1.88e+00	2.37e+00	0.599	0.771	0.775	0.575	0.770	0.774	0.01450	0.372	5.21	4.39	0.20
ADAS11	0.02930	1.02e+00	1.03e+00	1.04e+00	0.676	0.714	0.740	0.667	0.718	0.739	0.04093	0.389	5.09	4.59	0.65
M_ST29SV	-18.49681	7.10e-12	9.27e-09	1.21e-05	0.666	0.751	0.756	0.671	0.748	0.757	0.02289	0.368	4.98	4.30	0.95
M_ST13TA	-21.98696	6.17e-14	2.83e-10	1.30e-06	0.669	0.725	0.736	0.667	0.729	0.739	0.03067	0.371	4.92	4.37	0.20
RD_ST25TA	-0.30414	6.49e-01	7.38e-01	8.39e-01	0.494	0.775	0.775	0.511	0.775	0.776	0.00806	0.254	4.45	2.97	0.20
M_ST40SA	-0.36302	5.88e-01	6.96e-01	8.22e-01	0.632	0.713	0.732	0.627	0.714	0.733	0.01452	0.307	4.10	3.55	0.10
M_ST51TS	89.89985	5.20e+20	1.10e+39	2.35e+57	0.553	0.748	0.755	0.553	0.750	0.757	0.01205	0.348	4.08	4.05	0.45
RAVLT_immediate	-0.02411	9.65e-01	9.76e-01	9.87e-01	0.658	0.781	0.772	0.677	0.781	0.774	0.01950	0.456	4.06	5.39	1.00
M_ST24CV	-6.28493	1.02e-04	1.86e-03	3.41e-02	0.703	0.749	0.752	0.689	0.749	0.752	0.01131	0.256	4.02	2.99	0.60
RAVLT_perc_forgetting	0.00268	1.00e+00	1.00e+00	1.00e+00	0.638	0.736	0.740	0.652	0.730	0.739	0.02211	0.407	4.00	4.80	0.65
M_ST43TS	95.08098	1.37e+21	1.96e+41	2.81e+61	0.541	0.751	0.756	0.535	0.752	0.757	0.01060	0.299	3.84	3.47	0.60
M_ST40CV	-7.85417	7.44e-06	3.88e-04	2.02e-02	0.699	0.741	0.755	0.682	0.741	0.755	0.01736	0.199	3.81	2.30	0.50
M_ST39SA	-1.25517	1.55e-01	2.85e-01	5.24e-01	0.631	0.766	0.769	0.619	0.765	0.769	0.00993	0.366	3.78	4.29	0.35
MMSE	-0.01820	9.73e-01	9.82e-01	9.91e-01	0.649	0.724	0.734	0.642	0.729	0.736	0.02320	0.423	3.68	5.00	0.20
RD_ST34CV	-0.95013	2.33e-01	3.87e-01	6.41e-01	0.533	0.730	0.742	0.555	0.733	0.745	0.01588	0.332	3.61	3.89	0.20
RD_ST129TS	-0.13012	8.19e-01	8.78e-01	9.41e-01	0.511	0.728	0.738	0.532	0.729	0.737	0.01408	0.264	3.61	3.09	0.10

pander::pander(bConvml$univariate[bConvml$selectedfeatures,])

	Name	RName	ZUni
ADAS13	ADAS13	ADAS13	10.41
FAQ	FAQ	FAQ	8.69
RAVLT_immediate	RAVLT_immediate	RAVLT_immediate	9.01
RD_ST24CV	RD_ST24CV	RD_ST24CV	3.38
RD_ST25TA	RD_ST25TA	RD_ST25TA	1.45
M_ST39SA	M_ST39SA	M_ST39SA	5.26
M_ST13CV	M_ST13CV	M_ST13CV	6.56
ADAS11	ADAS11	ADAS11	8.86
M_ST24CV	M_ST24CV	M_ST24CV	9.05
M_ST29SV	M_ST29SV	M_ST29SV	8.83
M_ST40CV	M_ST40CV	M_ST40CV	8.80
M_ST43TS	M_ST43TS	M_ST43TS	1.83
RAVLT_perc_forgetting	RAVLT_perc_forgetting	RAVLT_perc_forgetting	6.63
M_ST51TS	M_ST51TS	M_ST51TS	2.57
RD_ST49CV	RD_ST49CV	RD_ST49CV	1.87
M_ST40SA	M_ST40SA	M_ST40SA	5.80
M_ST31CV	M_ST31CV	M_ST31CV	7.34
M_ST12SV	M_ST12SV	M_ST12SV	8.86
M_ST13TA	M_ST13TA	M_ST13TA	7.36
RD_ST129TS	RD_ST129TS	RD_ST129TS	1.83
MMSE	MMSE	MMSE	5.83
RD_ST43TS	RD_ST43TS	RD_ST43TS	3.01
RD_ST34CV	RD_ST34CV	RD_ST34CV	1.79

ptestl <- predict(bConvml,TADPOLE_Conv_TEST,type="lp")
boxplot(ptestl~TADPOLE_Conv_TEST$status)

ptestr <- predict(bConvml,TADPOLE_Conv_TEST,type="risk")
eventCases <- subset(TADPOLE_Conv_TEST,status==1)
plot(1.0/ptestr[rownames(eventCases)]~eventCases$TimeToEvent)

pander::pander(cor.test(eventCases$TimeToEvent,1.0/ptestr[rownames(eventCases)],method="spearman"))

Spearman’s rank correlation rho: eventCases$TimeToEvent and 1/ptestr[rownames(eventCases)]

Test statistic	P value	Alternative hypothesis	rho
91411	1.92e-05 * * *	two.sided	0.417

perdsurv <- cbind(TADPOLE_Conv_TEST$TimeToEvent,
                  TADPOLE_Conv_TEST$status,
                  ptestl,
                  ptestr)
prSurv <- predictionStats_survival(perdsurv,"MCI to  AD Conversion")

pander::pander(prSurv$CIRisk)

median	lower	upper
0.835	0.799	0.871

pander::pander(prSurv$CILp)

median	lower	upper
0.852	0.804	0.894

pander::pander(prSurv$spearmanCI)

50%	2.5%	97.5%
0.418	0.231	0.581

prBin <- predictionStats_binary(cbind(TADPOLE_Conv_TEST$status,ptestl),"MCI to  AD Conversion")

MCI to AD Conversion

pander::pander(prBin$aucs)

est	lower	upper
0.852	0.807	0.897

pander::pander(prBin$CM.analysis$tab)

	Outcome +	Outcome -	Total
Test +	76	47	123
Test -	22	128	150
Total	98	175	273

par(op)

5.1.2 The formula network

cmax <- apply(bConvml$bagging$formulaNetwork,2,max)
cnames <- names(cmax[cmax>=0.25])
cmax <- cmax[cnames]

adma <- bConvml$bagging$formulaNetwork[cnames,cnames]

adma[adma<0.15] <- 0;
gr <- graph_from_adjacency_matrix(adma,mode = "undirected",diag = FALSE,weighted=TRUE)
gr$layout <- layout_with_fr

fc <- cluster_optimal(gr)
plot(fc, gr,
     edge.width=5*E(gr)$weight,
     vertex.size=20*cmax,
     vertex.label.cex=0.75,
     vertex.label.dist=0,
     main="MCI to Dementia Conversion")

par(op)

5.1.3 MCI to Dementia table

clusterFeatures <- fc$names
sm$coefficients$DeltaAUC <- (sm$coefficients$full.AUC-sm$coefficients$r.AUC)

tableMCI_to_Dem <- sm$coefficients[clusterFeatures,
                                   c("Estimate",
                                     "lower",
                                     "HR",
                                     "upper",
                                     "full.AUC",
                                     "DeltaAUC",
                                     "z.IDI",
                                     "Frequency")]

nugget <- fc$membership
names(nugget) <- clusterFeatures

tableMCI_to_Dem$Cluster <- nugget[rownames(tableMCI_to_Dem)]

rnames <- clusterFeatures[str_detect(clusterFeatures,"ST")]
frnames <- rnames
rnames <- str_replace_all(rnames,"M_","")
rnames <- str_replace_all(rnames,"RD_","")
rnames <- str_replace_all(rnames,"Ba_","")
rnames <- str_replace_all(rnames,"De_","")
description <- character()

for (ddet in c(1:length(rnames)))
{
  description <- c(description,TADPOLE_D1_D2_Dict$TEXT[str_detect(TADPOLE_D1_D2_Dict$FLDNAME,rnames[ddet])][1])
}
names(description) <- frnames

tableMCI_to_Dem$Description <- description[rownames(tableMCI_to_Dem)]
pander::pander(tableMCI_to_Dem)

	Estimate	lower	HR	upper	full.AUC	DeltaAUC	z.IDI	Frequency	Cluster	Description
ADAS13	0.06511	1.05e+00	1.07e+00	1.09e+00	0.774	0.01847	5.86	1.00	1	NA
FAQ	0.06129	1.05e+00	1.06e+00	1.08e+00	0.774	0.03236	6.51	1.00	1	NA
RAVLT_immediate	-0.02411	9.65e-01	9.76e-01	9.87e-01	0.774	-0.00720	4.06	1.00	1	NA
M_ST39SA	-1.25517	1.55e-01	2.85e-01	5.24e-01	0.769	0.00427	3.78	0.35	1	Surface Area of LeftMedialOrbitofrontal
ADAS11	0.02930	1.02e+00	1.03e+00	1.04e+00	0.739	0.02100	5.09	0.65	2	NA
M_ST24CV	-6.28493	1.02e-04	1.86e-03	3.41e-02	0.752	0.00366	4.02	0.60	1	Volume (Cortical Parcellation) of LeftEntorhinal
M_ST29SV	-18.49681	7.10e-12	9.27e-09	1.21e-05	0.757	0.00955	4.98	0.95	2	Volume (WM Parcellation) of LeftHippocampus
M_ST40CV	-7.85417	7.44e-06	3.88e-04	2.02e-02	0.755	0.01423	3.81	0.50	2	Volume (Cortical Parcellation) of LeftMiddleTemporal
M_ST43TS	95.08098	1.37e+21	1.96e+41	2.81e+61	0.757	0.00458	3.84	0.60	2	Cortical Thickness Standard Deviation of LeftParacentral
RAVLT_perc_forgetting	0.00268	1.00e+00	1.00e+00	1.00e+00	0.739	0.00901	4.00	0.65	2	NA
M_ST51TS	89.89985	5.20e+20	1.10e+39	2.35e+57	0.757	0.00694	4.08	0.45	2	Cortical Thickness Standard Deviation of LeftPrecentral

5.2 Decorrelated

TADPOLE_Conv_TRAIND <- GDSTMDecorrelation(TADPOLE_Conv_TRAIN,
                                          Outcome="status",
                                          thr=0.6,
                                          type="RLM",
                                          method="spearman",
                                          verbose = TRUE)
#> 
#>  Included: 203 , Uni p: 0.01172495 To Outcome: 103 , Base: 3 , In Included: 3 , Base Cor: 3 
#> 1 , Top: 45 < 0.6 >( 2 )[ 1 : 0 : 0.594 ]( 43 , 98 , 0 ),<>Tot Used: 141 , Added: 98 , Zero Std: 0 , Max Cor: 0.9507145 
#> 2 , Top: 24 < 0.6 >[ FALSE ]( 1 )[ 1 : 0 : 0.594 ]( 22 , 35 , 43 ),<>Tot Used: 162 , Added: 35 , Zero Std: 0 , Max Cor: 0.8978966 
#> 3 , Top: 22 < 0.6 >( 2 )[ 1 : 0 : 0 ]( 21 , 25 , 62 ),<>Tot Used: 168 , Added: 25 , Zero Std: 0 , Max Cor: 0.9199215 
#> 4 , Top: 9 < 0.6 >[ FALSE ]( 1 )[ 1 : 0 : 0 ]( 9 , 9 , 77 ),<>Tot Used: 168 , Added: 9 , Zero Std: 0 , Max Cor: 0.8275661 
#> 5 , Top: 1 < 0.6 >[ FALSE ]( 1 )[ 1 : 0 : 0.6 ]( 1 , 1 , 83 ),<>Tot Used: 169 , Added: 1 , Zero Std: 0 , Max Cor: 0.5981195 
#> [ 6 ], 0.5952519 . Cor to Base: 107 , ABase: 69
TADPOLE_Conv_TESTD <-  predictDecorrelate(TADPOLE_Conv_TRAIND,TADPOLE_Conv_TEST)

5.2.1 Learning

bConvmlD <- BSWiMS.model(Surv(TimeToEvent,status)~1,TADPOLE_Conv_TRAIND,NumberofRepeats = 20)

[+–+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-]..

pander::pander(bConvmlD$bagging$Jaccard.SM)

0.579

fs <- bConvmlD$bagging$frequencyTable
barplot(fs[order(-fs)],las=2,main="Selected Features",cex.names = 0.5)

sm <- summary(bConvmlD)
pander::pander(sm$coefficients)

	Estimate	lower	HR	upper	u.Accuracy	r.Accuracy	full.Accuracy	u.AUC	r.AUC	full.AUC	IDI	NRI	z.IDI	z.NRI	Frequency
Ba_ADAS13	0.1230	1.11e+00	1.13e+00	1.16e+00	0.702	0.741	0.772	0.698	0.737	0.774	0.09878	0.6188	8.362	7.466	1.00
FAQ	0.0999	1.08e+00	1.11e+00	1.13e+00	0.732	0.738	0.772	0.702	0.741	0.774	0.04870	0.5739	7.443	7.154	1.00
Ba_Hippocampus	-29.1994	1.80e-18	2.08e-13	2.41e-08	0.660	0.767	0.772	0.667	0.764	0.774	0.02331	0.4220	4.794	4.952	0.95
RD_ST24CV	0.1424	1.09e+00	1.15e+00	1.22e+00	0.600	0.768	0.773	0.575	0.763	0.770	0.00878	0.2682	4.560	3.130	0.05
De_M_ST13TA	-135.3800	3.95e-85	1.60e-59	6.51e-34	0.609	0.769	0.776	0.606	0.774	0.778	0.02280	0.4613	4.426	5.436	0.50
De_M_ST25CV	60.8165	1.60e+15	2.58e+26	4.16e+37	0.554	0.777	0.774	0.550	0.778	0.776	0.01309	0.2641	4.365	3.066	0.65
De_ADAS11	-0.0347	9.51e-01	9.66e-01	9.81e-01	0.543	0.785	0.779	0.547	0.782	0.781	0.01292	0.4015	4.221	4.703	0.25
De_M_ST51TS	111.1803	1.83e+25	1.93e+48	2.04e+71	0.498	0.770	0.767	0.498	0.773	0.770	0.01086	0.2769	4.020	3.208	0.25
De_M_ST39CV	-1.3250	1.33e-01	2.66e-01	5.30e-01	0.641	0.773	0.773	0.634	0.765	0.770	0.01408	0.3244	3.450	3.782	0.05
De_RAVLT_immediate	-0.0136	9.79e-01	9.86e-01	9.94e-01	0.558	0.781	0.770	0.560	0.781	0.773	0.01596	0.4163	3.412	4.922	0.45
M_ST43TS	80.5135	6.91e+15	9.26e+34	1.24e+54	0.541	0.772	0.773	0.536	0.770	0.774	0.00982	0.2967	3.375	3.443	0.30
RD_ST25TA	-0.0641	8.83e-01	9.38e-01	9.96e-01	0.500	0.773	0.773	0.497	0.769	0.770	0.00372	-0.0102	0.382	-0.127	0.05

pander::pander(bConvmlD$univariate[bConvmlD$selectedfeatures,])

	Name	RName	ZUni
Ba_ADAS13	Ba_ADAS13	Ba_ADAS13	10.41
FAQ	FAQ	FAQ	8.69
Ba_Hippocampus	Ba_Hippocampus	Ba_Hippocampus	8.46
RD_ST24CV	RD_ST24CV	RD_ST24CV	3.38
De_M_ST39CV	De_M_ST39CV	De_M_ST39CV	6.01
De_M_ST51TS	De_M_ST51TS	De_M_ST51TS	1.53
De_RAVLT_immediate	De_RAVLT_immediate	De_RAVLT_immediate	2.45
De_ADAS11	De_ADAS11	De_ADAS11	1.50
De_M_ST25CV	De_M_ST25CV	De_M_ST25CV	4.06
RD_ST25TA	RD_ST25TA	RD_ST25TA	1.45
M_ST43TS	M_ST43TS	M_ST43TS	1.83
De_M_ST13TA	De_M_ST13TA	De_M_ST13TA	4.51

ptestl <- predict(bConvmlD,TADPOLE_Conv_TESTD,type="lp")
boxplot(ptestl~TADPOLE_Conv_TEST$status)

ptestr <- predict(bConvmlD,TADPOLE_Conv_TESTD,type="risk")
eventCases <- subset(TADPOLE_Conv_TEST,status==1)
plot(1.0/ptestr[rownames(eventCases)]~eventCases$TimeToEvent)

pander::pander(cor.test(eventCases$TimeToEvent,1.0/ptestr[rownames(eventCases)],method="spearman"))

Spearman’s rank correlation rho: eventCases$TimeToEvent and 1/ptestr[rownames(eventCases)]

Test statistic	P value	Alternative hypothesis	rho
87071	4.44e-06 * * *	two.sided	0.445

perdsurv <- cbind(TADPOLE_Conv_TEST$TimeToEvent,
                  TADPOLE_Conv_TEST$status,
                  ptestl,
                  ptestr)
prSurv <- predictionStats_survival(perdsurv,"MCI to  AD Conversion")

pander::pander(prSurv$CIRisk)

median	lower	upper
0.839	0.803	0.871

pander::pander(prSurv$CILp)

median	lower	upper
0.854	0.807	0.896

pander::pander(prSurv$spearmanCI)

50%	2.5%	97.5%
0.443	0.262	0.603

prBin <- predictionStats_binary(cbind(TADPOLE_Conv_TESTD$status,ptestl),"MCI to  AD Conversion")

MCI to AD Conversion

pander::pander(prBin$aucs)

est	lower	upper
0.854	0.809	0.899

pander::pander(prBin$CM.analysis$tab)

	Outcome +	Outcome -	Total
Test +	80	46	126
Test -	18	129	147
Total	98	175	273

par(op)

5.2.2 The formula network

cmax <- apply(bConvmlD$bagging$formulaNetwork,2,max)
cnames <- names(cmax[cmax>=0.25])
cmax <- cmax[cnames]

adma <- bConvmlD$bagging$formulaNetwork[cnames,cnames]

adma[adma<0.15] <- 0;
gr <- graph_from_adjacency_matrix(adma,mode = "undirected",diag = FALSE,weighted=TRUE)
gr$layout <- layout_with_fr

fc <- cluster_optimal(gr)
plot(fc, gr,
     edge.width=5*E(gr)$weight,
     vertex.size=20*cmax,
     vertex.label.cex=0.75,
     vertex.label.dist=0,
     main="MCI to Dementia Conversion")

par(op)

5.2.3 Decorrelated MCI to Dementia table

clusterFeatures <- fc$names
sm$coefficients$DeltaAUC <- (sm$coefficients$full.AUC-sm$coefficients$r.AUC)

tableMCI_to_DemD <- sm$coefficients[clusterFeatures,
                                   c("Estimate",
                                     "lower",
                                     "HR",
                                     "upper",
                                     "full.AUC",
                                     "DeltaAUC",
                                     "z.IDI",
                                     "Frequency")]

nugget <- fc$membership
names(nugget) <- clusterFeatures

tableMCI_to_DemD$Cluster <- nugget[rownames(tableMCI_to_DemD)]

rnames <- clusterFeatures[str_detect(clusterFeatures,"ST")]
frnames <- rnames
rnames <- str_replace_all(rnames,"M_","")
rnames <- str_replace_all(rnames,"RD_","")
rnames <- str_replace_all(rnames,"Ba_","")
rnames <- str_replace_all(rnames,"De_","")
description <- character()

for (ddet in c(1:length(rnames)))
{
  description <- c(description,TADPOLE_D1_D2_Dict$TEXT[str_detect(TADPOLE_D1_D2_Dict$FLDNAME,rnames[ddet])][1])
}
names(description) <- frnames

tableMCI_to_DemD$Description <- description[rownames(tableMCI_to_DemD)]


## Getting the decorrelation formula
dc <- getDerivedCoefficients(TADPOLE_Conv_TRAIND)
decornames <- rownames(sm$coefficients)

deNames_in_dc <- decornames[decornames %in% names(dc)]
theDeFormulas <- dc[deNames_in_dc]
deFromula <- character(length(theDeFormulas))
names(deFromula) <- names(theDeFormulas)
for (dx in names(deFromula))
{
  coef <- theDeFormulas[[dx]]
  cname <- names(theDeFormulas[[dx]])
  names(cname) <- cname
  for (cf in names(coef))
  {
    if (cf != dx)
    {
      if (coef[cf]>0)
      {
        deFromula[dx] <- paste(deFromula[dx],
                               sprintf("+ %5.3f*%s",coef[cf],cname[cf]))
      }
      else
      {
        deFromula[dx] <- paste(deFromula[dx],
                               sprintf("%5.3f*%s",coef[cf],cname[cf]))
      }
    }
  }
}

tableMCI_to_DemD$DecorFormula <- deFromula[rownames(tableMCI_to_DemD)]


pander::pander(tableMCI_to_DemD)

	Estimate	lower	HR	upper	full.AUC	DeltaAUC	z.IDI	Frequency	Cluster	Description	DecorFormula
Ba_ADAS13	0.1230	1.11e+00	1.13e+00	1.16e+00	0.774	0.037173	8.36	1.00	1	NA	NA
FAQ	0.0999	1.08e+00	1.11e+00	1.13e+00	0.774	0.033366	7.44	1.00	1	NA	NA
Ba_Hippocampus	-29.1994	1.80e-18	2.08e-13	2.41e-08	0.774	0.009979	4.79	0.95	1	NA	NA
De_M_ST51TS	111.1803	1.83e+25	1.93e+48	2.04e+71	0.770	-0.003077	4.02	0.25	1	Cortical Thickness Standard Deviation of LeftPrecentral	-0.605M_ST15TS + 1.000M_ST51TS
De_RAVLT_immediate	-0.0136	9.79e-01	9.86e-01	9.94e-01	0.773	-0.007865	3.41	0.45	1	NA	+ 0.971ADAS13 + 1.000RAVLT_immediate
De_ADAS11	-0.0347	9.51e-01	9.66e-01	9.81e-01	0.781	-0.000775	4.22	0.25	1	NA	+ 1.000ADAS11 -0.643ADAS13
De_M_ST25CV	60.8165	1.60e+15	2.58e+26	4.16e+37	0.776	-0.001914	4.36	0.65	1	Volume (Cortical Parcellation) of LeftFrontalPole	-1.088M_ST25TA -0.287M_ST25SA + 1.000M_ST25CV -0.262M_ST39CV
M_ST43TS	80.5135	6.91e+15	9.26e+34	1.24e+54	0.774	0.003489	3.38	0.30	1	Cortical Thickness Standard Deviation of LeftParacentral	NA
De_M_ST13TA	-135.3800	3.95e-85	1.60e-59	6.51e-34	0.778	0.003874	4.43	0.50	1	Cortical Thickness Average of LeftBankssts	+ 1.000M_ST13TA -0.892M_ST59TA

5.2.4 Saving the enviroment

save.image("~/GitHub/BSWiMS/TADPOLE_BSWIMS_Results.RData")